System

Abstract

To provide a system.SOLUTION: A system includes first means for receiving an inquiry from a user, second means for generating a first prompt sentence for outputting a first response to the inquiry based on the received inquiry from the user, third means for inputting the generated first prompt sentence into a generative AI model to generate the first response, fourth means for calculating a reliability score for the generated first response, and fifth means for displaying the generated first response and the calculated reliability score on a terminal of a person in charge.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] There is an issue where the front desk staff or operators cannot think of an immediate response to a complaint. This can lead to the complaint escalating and prolonging the response time. In addition, responding by making one's own judgment or consulting with a superior places a heavy mental burden, leading to inconsistencies in response skills and issues with continued employment. [Means for solving the problem]

[0005] This invention is a complaint handling support system for customer service representatives and operators, and provides a means to provide appropriate responses to complaints, prevent further escalation, and shorten response times. It also provides a means to provide psychological security and to level out response skills, thereby helping to resolve the issue of continued employment. [Brief explanation of the drawings]

[0006] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 2 is a sequence diagram showing a flow of processing in the data processing system according to the first embodiment of the first form example. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Embodiment 1. [Figure 13] FIG. 10 is a sequence diagram showing a processing flow of a data processing system in a second embodiment of the second form example. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 of Embodiment Example 2. [Figure 15]FIG. 10 is a sequence diagram showing the flow of processing in a data processing system according to a third embodiment of the third embodiment. [Figure 16] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 3 of Embodiment 3. [Figure 17] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the first embodiment of the first form example when an emotion engine is combined. [Figure 18] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Form Example 1 when an emotion engine is combined. [Figure 19] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the second embodiment of the second form example when an emotion engine is combined. [Figure 20] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 of Form Example 2 when an emotion engine is combined. [Figure 21] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the third embodiment of the third form example when an emotion engine is combined. [Figure 22] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 3 of Form Example 3 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0007] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0008] First, the terms used in the following description will be explained.

[0009] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, the processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), or a TPU (TENSOR PROCESSING UNIT (registered trademark)).

[0010] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0011] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0012] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0013] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0014] [First embodiment]

[0015] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0016] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0017] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0018] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0019] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0020] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0021] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0022] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0023] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0024] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0025] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0026] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0027] "Example 1"

[0028] A first embodiment of the present invention is a complaint handling support system for front desk staff and operators. This system is equipped with an automatic response generation module using AI (artificial intelligence) as a means of providing appropriate responses to complaints. This module analyzes the content of the complaint and generates an appropriate response. It also has a function to predict the escalation of the complaint as a means of preventing further escalation. Furthermore, as a means of shortening response time, an optimization algorithm is used to speed up response generation. "Example 2"

[0029] A second embodiment of the present invention is a system further including a means for providing a sense of mental security and a means for achieving the standardization of response skills. As a means for providing a sense of mental security, the system is provided with a function for displaying a reliability score for the response generated by the AI. This allows the operator to confirm the reliability of the response proposed by the AI ​​and achieve a sense of mental security. Furthermore, as a means for achieving the standardization of response skills, the operator can refer to the response generated by the AI, thereby reducing the variation in response skills between operators.

[0030] "Example 3"

[0031] A third embodiment of the present invention is a system that further includes a means for supporting the resolution of the issue of continued employment. Specifically, the system is equipped with a function for evaluating the quality of responses generated by AI and supporting the improvement of the operator's skills. This helps improve the operator's skills and maintain their motivation, thereby supporting the resolution of the issue of continued employment.

[0032] The processing flow of each embodiment will be described below.

[0033] "Example 1"

[0034] Step 1: The counter clerk or operator receives the complaint.

[0035] Step 2: The system analyzes the content of the claim.

[0036] Step 3: An automatic response generation module using AI (artificial intelligence) generates an appropriate response.

[0037] Step 4: The system predicts claim escalation.

[0038] Step 5: An optimization algorithm is implemented to speed up response generation, reducing response time.

[0039] "Example 2"

[0040] Step 1: The teller or operator reviews the response generated by the AI.

[0041] Step 2: The system displays a confidence score for the AI-generated response.

[0042] Step 3: The teller or operator decides on the final response based on the reliability score.

[0043] Step 4: Counter staff and operators use the responses generated by the AI ​​as a reference to standardize their response skills.

[0044] "Example 3"

[0045] Step 1: The teller / operator uses the AI-generated response.

[0046] Step 2: The system evaluates the quality of the AI-generated response.

[0047] Step 3: The system provides feedback to help operators improve their skills.

[0048] Step 4: Counter staff and operators use the feedback to improve their skills and maintain motivation.

[0049] Example 1

[0050] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0051] Conventional complaint handling systems did not provide sufficient support for the front desk staff or operators to respond appropriately to complaints, which often resulted in variations in the quality and speed of responses. In addition, there was no way to predict the risk of a complaint escalating, and a delayed response was likely to escalate the situation. Furthermore, with the need to reduce response times, it was difficult to generate a prompt and appropriate response. A system that could solve these issues was needed.

[0052] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0053] In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further escalation, means for shortening response time, means for analyzing the content of the complaint, means for generating automatic responses, means for predicting escalation, and means for accelerating response generation. This improves the quality and speed of complaint handling, reduces the risk of escalation, and makes it possible to provide prompt and appropriate responses.

[0054] The "means for providing an appropriate response to a complaint" is a function for generating and providing an appropriate response to a complaint from a user.

[0055] "Measures to prevent further escalation" is a function to prevent complaints from escalating and the problem from expanding.

[0056] The "means for realizing a reduction in response time" is a function for reducing response time by quickly generating and providing responses to complaints.

[0057] The "means for analyzing the content of a complaint" is a function for analyzing the content of a complaint from a user and identifying the type of problem and the emotion behind it.

[0058] The "means for generating an automatic response" is a function for automatically generating an appropriate response based on the analysis results.

[0059] "Means for predicting escalation" is a function for predicting the possibility of a claim escalating and assessing the risk.

[0060] The "means for speeding up response generation" is a function for optimizing the response generation process and providing a response quickly.

[0061] This invention is a support system for enabling counter staff and operators to respond appropriately to complaints. This system includes multiple means for providing appropriate responses to complaints, preventing flaming, and shortening response times.

[0062] System configuration

[0063] Hardware and Software

[0064] Server: Analyzes complaint content, generates automatic responses, predicts escalations, and speeds up response generation.

[0065] Terminal: The device (PC, smartphone, etc.) through which the user enters claims and receives responses from the server.

[0066] Natural language processing (NLP) technology: Used to analyze the content of claims. Specifically, Google® Cloud Natural Language API and IBM Watson® Natural Language Understanding® are used.

[0067] Generative AI model: Used to generate automated responses. Specifically, we use OpenAI's GPT-3.

[0068] Machine learning algorithms: Used to predict escalation. Specifically, we use scikit-learn (registered trademark) and TENSORFLOW (registered trademark).

[0069] Caching techniques and parallel processing are used to speed up response generation.

[0070] System Operation

[0071] 1. Entering a claim

[0072] The user inputs the details of the complaint from the terminal.

[0073] Example: User types "The product is of poor quality."

[0074] 2. Analysis of the complaint

[0075] The server receives the complaint sent by the user.

[0076] The server parses the claim content using the Google Cloud Natural Language API.

[0077] Specifically, the text of the complaint is sent to an API, which performs sentiment analysis and keyword extraction on the text.

[0078] Example: Extract the keyword "quality issue" from the text "poor product quality" and determine the sentiment as "negative."

[0079] 3. Generate Auto-Responses

[0080] The server generates an appropriate response using a generative AI model (GPT-3) based on the analysis results.

[0081] Specifically, the analysis results are input into a generative AI model as a prompt sentence, and a response sentence is generated.

[0082] Example: Enter the prompt text "The user has entered a complaint that the product quality is poor. Please generate an appropriate response." and get the generated response "I'm sorry. Can you tell me more about the quality of the product?"

[0083] 4. Escalation Prediction

[0084] The server uses machine learning algorithms to predict the likelihood of a claim escalation.

[0085] Specifically, we use scikit-learn to build a model that evaluates the risk of escalation based on past claim data and calculate a risk score for current claims.

[0086] Example: A high risk score is assigned to a complaint that has not been resolved despite repeated contact.

[0087] 5. Providing a Response

[0088] The server provides the generated response to the user.

[0089] Specifically, the generated response sentence is sent to the terminal and displayed on the user's screen.

[0090] Example: "Sorry, can you tell me more about the quality of your product?" is the response displayed on the user's device.

[0091] 6. Faster response generation

[0092] The server uses caching techniques and parallel processing to speed up response generation.

[0093] Specifically, past responses to the same type of claim are cached and reused, and processing speed is improved by processing multiple claims in parallel.

[0094] Example: If the response to a complaint about "poor quality of product" is stored in the cache, retrieve the response from the cache and serve it quickly.

[0095] Specific examples

[0096] The user inputs a complaint saying that the product is of poor quality.

[0097] The server sends the complaint content to the Google Cloud Natural Language API and extracts the keyword "quality issue" and the sentiment "negative."

[0098] The server inputs the prompt text "The user has input a complaint that 'the quality of the product is poor.' Please generate an appropriate response." into the generative AI model (GPT-3), which generates the response "I'm sorry. Could you tell me more about the quality of the product?"

[0099] The server calculates the risk score for escalation using scikit-learn and assigns a high risk score.

[0100] The server sends the generated response to the user's terminal and displays it on the user's screen.

[0101] The server caches responses to the same type of claim to speed up subsequent response generation.

[0102] The above is an embodiment of the present invention.

[0103] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0104] Step 1:

[0105] Entering a claim

[0106] The user inputs the details of the complaint from the terminal.

[0107] Input: The complaint entered by the user (e.g., "The product is of poor quality").

[0108] Output: The claim is sent to the server.

[0109] Specific operation: The user enters the details of the complaint into the input form on the terminal and presses the send button.

[0110] Step 2:

[0111] Receiving complaint details

[0112] The server receives the complaint sent by the user.

[0113] Input: The complaint sent by the user.

[0114] Output: The claim is saved on the server.

[0115] Specific operation: The server receives the HTTP request and saves the claim details in the database.

[0116] Step 3:

[0117] Analysis of complaints

[0118] The server parses the claim content using the Google Cloud Natural Language API.

[0119] Input: Saved claim details.

[0120] Output: Analysis results (keywords, sentiment analysis results, etc.).

[0121] Specific operation: The server sends the claim content to the API and obtains the analysis results returned by the API.

[0122] Step 4:

[0123] Generate an auto-response

[0124] The server generates an appropriate response using a generative AI model (GPT-3) based on the analysis results.

[0125] Input: Analysis results.

[0126] Output: The generated response sentence.

[0127] Specific operation: The server inputs the analysis result as a prompt sentence into the generative AI model and obtains the generated response sentence.

[0128] Example: Enter the prompt text "The user has entered a complaint that the product quality is poor. Please generate an appropriate response." to generate the response "I'm sorry. Can you tell me more about the quality of your product?"

[0129] Step 5:

[0130] Escalation Prediction

[0131] The server uses machine learning algorithms to predict the likelihood of a claim escalation.

[0132] Input: Claim details and analysis results.

[0133] Output: Escalation risk score.

[0134] What happens: The server uses a model built from past claims data to calculate a risk score for the current claim.

[0135] Example: A high risk score is assigned to a complaint that has not been resolved despite repeated contact.

[0136] Step 6:

[0137] Providing a response

[0138] The server provides the generated response to the user.

[0139] Input: The generated response sentence.

[0140] Output: The response is displayed on the user's terminal.

[0141] Specific operation: The server sends the generated response to the user's terminal and displays it on the user's screen.

[0142] Example: "Sorry, can you tell me more about the quality of your product?" is the response displayed on the user's device.

[0143] Step 7:

[0144] Faster response generation

[0145] The server uses caching techniques and parallel processing to speed up response generation.

[0146] Input: Past complaints and responses.

[0147] Output: The cached response.

[0148] What it does: The server caches and reuses past responses to the same type of claim, and processes multiple claims in parallel to improve processing speed.

[0149] Example: If the response to a complaint about "poor quality of product" is stored in the cache, retrieve the response from the cache and serve it quickly.

[0150] (Application example 1)

[0151] Next, a description will be given of Application Example 1 of Embodiment Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0152] With conventional complaint handling systems, customer service representatives and operators had to manually handle complaints, resulting in long response times and lower customer satisfaction. It was also difficult to predict the escalation of complaints, meaning that it was not possible to notify superiors at the appropriate time, increasing the risk of complaints escalating. Furthermore, there was a large variation in response skills and a significant mental burden, which led to issues with continued employment.

[0153] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0154] In this invention, the server includes a means for providing appropriate responses to complaints, a means for preventing further escalation, a means for shortening response time, a means for analyzing the content of the complaint and generating an appropriate response, a means for predicting the escalation of the complaint, a means for automatically notifying a superior, a means for generating responses using a generative AI model, and a means for generating responses using prompt sentences. This enables efficient and rapid complaint handling, improving customer satisfaction and reducing the risk of complaints escalating. It also solves the issue of continued employment by leveling out response skills and providing psychological security.

[0155] A "contact person" is a person in charge of directly responding to inquiries and complaints from customers.

[0156] An "operator" is a person who operates a system or machine and is responsible for carrying out a specific task.

[0157] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[0158] "Means for providing appropriate responses" refers to the function for generating and providing appropriate responses to complaints.

[0159] "Measures to prevent flame wars" are functions to prevent complaints from spreading and problems from becoming more serious.

[0160] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[0161] The "means for analyzing the content of the claim and generating an appropriate response" is a function for analyzing the content of the claim and generating an appropriate response based on that analysis.

[0162] The "means for predicting the escalation of a complaint" is a function for predicting the possibility that a complaint will be escalated to a higher-level person in charge.

[0163] The "means for automatically notifying a superior" is a function for automatically notifying a superior when a complaint escalates.

[0164] "Response generation means using a generative AI model" is a function for generating responses to claims using an artificial intelligence model.

[0165] The "means for generating a response using a prompt sentence" is a function for generating a response to a complaint using a specific input sentence (prompt sentence).

[0166] As an example of an embodiment of this invention, we will explain a customer support application for an online shopping site. This system uses AI to automatically generate appropriate responses to customer complaints and inquiries, enabling quick responses. It also has a function to predict complaint escalation and automatically notify higher-level personnel as necessary.

[0167] System Configuration

[0168] The system consists of the following major components:

[0169] 1. Complaint Analysis Module: Analyzes the content of the complaint and generates an appropriate response.

[0170] 2. Escalation prediction module: predicts the escalation of claims.

[0171] 3. Notification module: Automatically notify superiors when escalation is predicted.

[0172] 4. Generative AI model: An artificial intelligence model for generating responses to claims.

[0173] 5. Prompt generation module: Generates a response using a specific input sentence (prompt sentence).

[0174] Hardware and Software

[0175] The system uses smartphones or PCs with internet connectivity as hardware, and Python (registered trademark) and OpenAI API as software.

[0176] Processing flow

[0177] 1. Entering the details of the complaint: The user enters the details of the complaint.

[0178] 2. Response generation: The claim analysis module analyzes the content of the claim and generates an appropriate response using a generative AI model.

[0179] 3. Escalation Prediction: The escalation prediction module predicts the likelihood of a claim escalation.

[0180] 4. Notification: When an escalation is predicted, the notification module automatically notifies superiors.

[0181] Specific examples

[0182] For example, if a customer enters a complaint stating, "I did not receive the product, so I would like a refund," the system operates as follows:

[0183] 1. Entering the complaint details: The user enters, "I did not receive the product, so I would like a refund."

[0184] 2. Response generation: The complaint analysis module analyzes this content and uses a generative AI model to generate a response such as, "We apologize for the inconvenience. We will immediately investigate the issue of the non-delivery of the product and proceed with the refund process."

[0185] 3. Escalation Prediction: The escalation prediction module predicts that the likelihood of escalation for this claim is "high."

[0186] 4. Notification: The notification module automatically notifies the superior, who then takes over the response.

[0187] Prompt Sentence Examples

[0188] Please enter your complaint: I did not receive my item and would like a refund.

[0189] In this way, the efficiency of customer support on the online shopping site can be improved, and customer satisfaction can be improved.

[0190] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0191] Step 1:

[0192] The user inputs the complaint details.

[0193] Input: The user enters, "I never received my item and would like a refund."

[0194] Specific operation: The user enters the content of the complaint in text format into an input form on a smartphone or PC.

[0195] Step 2:

[0196] The server receives the complaint content and sends it to the complaint analysis module.

[0197] Input: The claim entered by the user.

[0198] Output: The claim content sent to the claim analysis module.

[0199] Specific operation: The server receives input from the user and sends the content to the complaint analysis module.

[0200] Step 3:

[0201] The claim analysis module analyzes the content of the claim and generates an appropriate response using a generative AI model.

[0202] Input: The claim content sent to the claim analysis module.

[0203] Output: The generated response text.

[0204] Specific operation: The claim analysis module analyzes the content of the claim using natural language processing technology and sends it as a prompt to the generative AI model. The generative AI model generates an appropriate response and returns it to the claim analysis module.

[0205] Step 4:

[0206] The server generates a response and sends it to the user.

[0207] Input: The generated response text.

[0208] Output: The response text that is displayed to the user.

[0209] Specific operation: The server receives the generated response and sends it to the user's device, which displays the response.

[0210] Step 5:

[0211] An escalation prediction module predicts the likelihood of a claim escalation.

[0212] Input: Claim details.

[0213] Output: Escalation prediction (high / low).

[0214] Specific operation: The escalation prediction module analyzes the content of the complaint and predicts the possibility of escalation using a generative AI model. The prediction result is returned to the server.

[0215] Step 6:

[0216] The server receives the predicted results of the escalation and automatically notifies the superior as necessary.

[0217] Input: The predicted outcome of the escalation.

[0218] Output: Notification to superiors.

[0219] Specific operation: The server receives the escalation prediction result, and if the prediction result is "high," it sends a notification to the superior via email or a dedicated application.

[0220] In this way, the system can make complaint handling more efficient and faster, improving customer satisfaction and reducing the risk of complaints escalating.

[0221] Example 2

[0222] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0223] Conventional complaint handling support systems do not provide sufficient support for operators to respond appropriately, and lack means to provide psychological security or to standardize response skills. They also lack effective means to shorten response times or prevent complaints from escalating. This increases the burden on operators and leads to inconsistent response quality, which is an issue.

[0224] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0225] In this invention, the server includes means for receiving an inquiry, means for generating a response using a generative AI model, means for calculating a reliability score for the generated response, means for displaying the response and the reliability score, means for providing an appropriate response to the complaint, means for preventing further flaming, and means for realizing a reduction in response time. This allows the operator to confirm the reliability of the response generated by the AI ​​and gain psychological peace of mind, while also reducing the variation in response skills, shortening response time, and preventing flaming.

[0226] The "means for receiving an inquiry" is a function that enables the server to receive an inquiry from a user.

[0227] "Means for generating a response using a generative AI model" is a function for generating an appropriate response to a received inquiry using a generative AI model.

[0228] The "means for calculating a confidence score for a generated response" is a function for evaluating the reliability of a generated response and calculating a confidence score.

[0229] The "means for displaying the response and the confidence score" is a function for visually displaying the generated response and its confidence score to the user.

[0230] "Means for providing appropriate responses to complaints" refers to a function for providing responses that allow appropriate responses to be made to complaints.

[0231] "Measures to prevent further escalation" is a function that provides preventative measures to prevent problems from escalating when dealing with complaints.

[0232] "Means to shorten response times" refers to functions that improve efficiency in order to shorten the time it takes to respond to complaints.

[0233] "Means for providing psychological reassurance" is a function that allows operators to gain psychological reassurance by confirming the reliability of the AI's responses.

[0234] "Means for achieving standardization of response skills" refers to a function for reducing the variation in response skills among operators and achieving uniform response.

[0235] The "means for supporting the resolution of issues regarding employment continuity" is a function for providing support for the resolution of issues regarding the employment continuity of operators.

[0236] This invention is a complaint handling support system that uses a generative AI model to generate responses and displays their reliability scores, thereby providing peace of mind to operators and achieving a leveling of response skills. Specific embodiments of this system are described below.

[0237] Server Processing

[0238] The server receives a query from a user. The query is a prompt sentence, such as "Please tell me the progress of the new project." The server inputs this query into a generative AI model (e.g., GPT-4 (registered trademark)) and generates an appropriate response. For the generated response, the server calculates a confidence score. This confidence score is calculated using the AI ​​model's internal evaluation mechanism or an additional evaluation algorithm. The server sends the generated response and confidence score to the terminal.

[0239] Terminal handling

[0240] The device receives the response and confidence score sent from the server. The received response and confidence score are displayed on the device screen. For example, it may say, "Current progress is going well. The next milestone is next week. (Confidence score: 85%)." The device makes it easy to understand visually by displaying a high confidence score in green and a low confidence score in red.

[0241] User Action

[0242] The user checks the response and the confidence score displayed on the device. The confidence score is used to evaluate the reliability of the response. For example, if the confidence score is 85%, the user can trust the response and adopt it. If the confidence score is low, the user can make a second inquiry. When making a second inquiry, the user can add more specific questions by referring to the previous response.

[0243] Specific examples

[0244] Example 1: Means of providing psychological security

[0245] The user sends a query such as "How is the progress of the new project going?" The server uses a generative AI model (e.g., GPT-4) to generate a response such as "Current progress is going well. The next milestone is next week," and calculates a confidence score of 85%. The device displays this response and the confidence score, and the user can check the confidence score and feel reassured.

[0246] Example 2: A means to achieve leveling of response skills

[0247] Operator A receives an inquiry saying, "What are the steps to take in handling a complaint?" The server uses a generative AI model to generate a response saying, "First, listen carefully to the customer, then confirm the details of the problem, and finally propose a solution." Operator A responds using this response as a reference, and Operator B can also refer to the same steps when handling similar inquiries, thereby reducing the variability in response skills.

[0248] Prompt Sentence Examples

[0249] "Please tell me how your new project is progressing."

[0250] "Please tell me the procedure for handling complaints."

[0251] These prompts can be fed into a generative AI model to generate an appropriate response and confidence score.

[0252] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0253] Step 1:

[0254] The server receives a query from a user. When the user inputs a prompt such as "Please tell me the progress of the new project" through the terminal and presses the send button, the server receives the query. The input is the user's query, and the output is the query data stored on the server.

[0255] Step 2:

[0256] The server inputs the received query into the generative AI model. Specifically, the server sends the query content to the generative AI model (e.g., GPT-4) via an API. The input is the query data, and the output is the prompt sent to the generative AI model.

[0257] Step 3:

[0258] The generative AI model generates a response based on the prompt sentence. The generative AI model uses an internal natural language processing algorithm to generate an appropriate response. The input is the prompt sentence, and the output is the generated response text.

[0259] Step 4:

[0260] The server calculates a confidence score for the generated response. The server evaluates the reliability of the response using the AI ​​model's internal evaluation mechanism or additional evaluation algorithms, and calculates the confidence score. The input is the generated response text, and the output is the confidence score.

[0261] Step 5:

[0262] The server sends the generated response and confidence score to the terminal. The server combines the response text and confidence score into a single data packet and sends it to the terminal. The input is the response text and confidence score, and the output is the data packet sent to the terminal.

[0263] Step 6:

[0264] The terminal receives the response and confidence score sent from the server. The terminal analyzes the received data packet and extracts the response text and confidence score. The input is the data packet sent from the server, and the output is the analyzed response text and confidence score.

[0265] Step 7:

[0266] The terminal displays the received response and confidence score to the user. The terminal screen displays "Current progress is good. The next milestone is next week. (Confidence score: 85%)." The input is the parsed response text and confidence score, and the output is the information displayed to the user.

[0267] Step 8:

[0268] The user checks the response and the confidence score displayed on the terminal. The user evaluates the reliability of the response based on the confidence score. The input is the displayed response text and the confidence score, and the output is the user's evaluation result.

[0269] Step 9:

[0270] The user can either adopt the response or ask again. If the confidence score is high, the user can trust the response and adopt it. If the confidence score is low, the user can ask again with more specific questions. The input is the user's evaluation result, and the output is the adoption of a new query or response.

[0271] (Application example 2)

[0272] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0273] In conventional complaint response support systems, operators had to provide appropriate responses to complaints, which placed a heavy mental burden on them, and their response skills often varied. Furthermore, when security operators received AI suggestions, they lacked the means to verify their reliability, making it difficult to respond quickly and appropriately. This resulted in issues such as longer response times and an increased risk of outrage.

[0274] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0275] In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further flaming, means for shortening response time, means for displaying a reliability score for the response generated by the AI, means for reducing the variability in response skills by having an operator refer to the response generated by the AI, and means for a security operator to receive the AI's suggestions and respond while checking the reliability score. This allows the operator to respond quickly and appropriately while gaining peace of mind.

[0276] A "contact person" is a person whose job is to directly respond to inquiries and complaints from customers and users.

[0277] An "operator" is a person whose job is to operate a system or machine and perform a specific task.

[0278] A "complaint handling support system" is a system that supports appropriate responses to complaints from customers and users.

[0279] "Means for providing appropriate responses" refers to the function of generating accurate and effective responses to complaints and providing them to operators.

[0280] "Measures to prevent flame wars" is a function that takes measures to prevent problems from escalating when dealing with complaints.

[0281] "Means for shortening response time" refers to a function for reducing the time it takes to respond to complaints.

[0282] "Means for displaying a reliability score for responses generated by AI" is a function that quantifies the reliability of responses generated by AI and displays it to the operator.

[0283] "Means to reduce variability in response skills by having operators refer to responses generated by AI" is a function that reduces the differences in response skills between operators by having operators refer to responses generated by AI.

[0284] A "security operator" is a person in charge of security-related tasks and whose job is to ensure the safety of systems and facilities.

[0285] A "confidence score" is a number used to evaluate the reliability of a response generated by AI.

[0286] As an embodiment of the present invention, a complaint handling support system used by a security operator will be specifically described.

[0287] System configuration

[0288] The system consists of the following major components:

[0289] 1. Server: Hosts the AI ​​model and generates responses to queries.

[0290] 2. Terminal: A device such as a smartphone or tablet used by a security operator.

[0291] 3. User interface: An application where operators can enter queries and see the AI's responses and confidence scores.

[0292] Hardware and software used

[0293] Hardware: smartphones, tablets, servers

[0294] Software: Flask (registered trademark) (a Python web framework), requests (a library for sending HTTP requests), generative AI models

[0295] Data processing and calculation

[0296] The server receives queries from security operators and sends prompts to the generative AI model, which generates responses to the queries and returns the responses to the server, which then calculates a confidence score for the responses and sends it to the operator's terminal.

[0297] Processing flow

[0298] 1. Entering a query: The security operator enters a query through the user interface of the terminal.

[0299] 2. Sending a query: The terminal sends a query to the server.

[0300] 3. Generate response: The server sends a query to the generative AI model and generates a response.

[0301] 4. Calculate confidence score: The server calculates a confidence score for the generated response.

[0302] 5. Displaying the response and score: The server sends the response and confidence score to the terminal for the operator to review.

[0303] Specific examples

[0304] For example, a security operator enters a query such as "What should we do if a suspicious person breaks into a building?" This query is sent to the server, and the generative AI model generates an appropriate response. The server calculates a confidence score for the generated response and displays the response and score on the operator's device.

[0305] Prompt Sentence Examples

[0306] What should you do if a suspicious person breaks into your building?

[0307] In this way, security operators can take prompt and appropriate action while referring to AI suggestions, which will provide peace of mind and reduce the variability in response skills.

[0308] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0309] Step 1:

[0310] A security operator enters a query through a terminal user interface.

[0311] Input: The query entered by the security operator (e.g., "What should we do if a suspicious person breaks into our building?")

[0312] Output: The query is entered into the terminal.

[0313] How it works: The security operator opens the application on their smartphone or tablet, enters a query in the text box, and presses the "send" button.

[0314] Step 2:

[0315] The device sends a query to the server.

[0316] Input: The query typed into the terminal

[0317] Output: A query is sent to the server.

[0318] Specific operation: The terminal application sends the entered query to the server as an HTTP request.

[0319] Step 3:

[0320] The server sends queries to the generative AI model and generates a response.

[0321] Input: The query sent to the server

[0322] Output: The response from the generative AI model

[0323] Specific operation: The server sends the received query as a prompt to the generative AI model, and the generative AI model generates a response to the query and returns it to the server.

[0324] Step 4:

[0325] The server calculates a confidence score for the generated response.

[0326] Input: Response from a generative AI model

[0327] Output: Confidence score

[0328] Specific operation: The server analyzes the content of the generated response and applies a confidence score calculation algorithm to calculate a confidence score.

[0329] Step 5:

[0330] The server sends the response and the confidence score to the device.

[0331] Input: Response and confidence score from the generative AI model

[0332] Output: Response and confidence score sent to the device

[0333] Specific operation: The server combines the response and the confidence score and sends it to the terminal as an HTTP response.

[0334] Step 6:

[0335] The terminal displays the response and confidence score to the security operator.

[0336] Input: The response and confidence score sent by the server

[0337] Output: Response and confidence score displayed to the security operator

[0338] Specific operation: The terminal application displays the received response and confidence score on the screen for the security operator to review.

[0339] Example 3

[0340] Next, a description will be given of a third embodiment of the third embodiment. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0341] Conventional complaint handling support systems did not provide enough support for counter staff and operators to respond appropriately, making it difficult to reduce response times, provide psychological security, and standardize response skills.In addition, it was difficult to improve the skills of operators and maintain their motivation, which led to issues with continuing employment.

[0342] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[0343] In this invention, the server includes a means for generating a response to a prompt using a generative AI model, a means for evaluating the quality of the generated response, and a means for providing feedback based on the evaluation results. This allows operators to respond appropriately and quickly, improving their skills and maintaining their motivation. It also provides a sense of security and standardizes response skills, resolving the issue of continued employment.

[0344] A "customer service representative" is an employee whose role is to directly respond to customer inquiries and complaints.

[0345] An "operator" is a person who operates the system and handles customer service and complaints.

[0346] A "complaint handling support system" is a system that provides support for appropriate responses to customer complaints.

[0347] A "generative AI model" is a model that uses artificial intelligence to generate a response to an input prompt.

[0348] A "prompt sentence" is text such as a question or instruction that a user enters into a system.

[0349] A "response" is a text answer or suggestion that a generative AI model generates based on a prompt.

[0350] A "quality assessment algorithm" is an algorithm for assessing the quality of a generated response.

[0351] "Feedback" refers to information provided to operators based on the evaluation results, encouraging them to improve their skills and improve.

[0352] "Skill improvement" refers to an operator improving their own abilities and techniques.

[0353] "Maintaining motivation" refers to operators maintaining their enthusiasm and motivation for their work.

[0354] "Mental security" refers to the sense of peace and security that an operator feels when performing their work.

[0355] "Leveling response skills" refers to ensuring that multiple operators have the same level of response skills.

[0356] "Continued employment" refers to an operator continuing to work at the same workplace for an extended period of time.

[0357] This invention is a support system for helping service representatives and operators respond appropriately to customer complaints. The system uses a generative AI model to generate responses to prompts, evaluates the quality of the responses, and provides feedback on the evaluation results to improve the skills of the operators and maintain their motivation.

[0358] System configuration

[0359] Hardware

[0360] Server: Provides the computational resources to run the generative AI model and quality assessment algorithm.

[0361] Terminal: A device through which a user can enter prompts and view responses and evaluation results from the server.

[0362] software

[0363] Generative AI models: For example, using natural language generation models such as GPT-4.

[0364] Quality assessment algorithm: For example, use a BERT®-based assessment model.

[0365] Program processing

[0366] The server receives the prompt sentence entered by the user from the device. It then generates a response to the prompt sentence using a generative AI model. The generated response is evaluated by a quality evaluation algorithm, and the evaluation result is sent to the device as feedback. The device then displays the evaluation result and the generated response to the user.

[0367] Specific examples

[0368] For example, if a user enters the prompt "What is the appropriate way to respond to a customer complaint?", the server will process it as follows:

[0369] 1. The server receives the prompt: "Please tell me the appropriate way to respond to a customer complaint."

[0370] 2. Use a generative AI model (e.g., GPT-4) to generate responses like this: "First, listen to the customer fully and demonstrate empathy. Then, offer a concrete solution and respond quickly."

[0371] 3. The server evaluates this response using a quality evaluation algorithm (e.g., a BERT-based evaluation model) and determines, for example, that "the quality of the response is high."

[0372] 4. The evaluation results are sent to the operator as feedback to help improve their skills.

[0373] Other examples of prompts include:

[0374] Please briefly explain the features of your new product.

[0375] "What is the proper way to respond to customer complaints?"

[0376] "Suggest ways to motivate your team."

[0377] This system allows operators to objectively evaluate the quality of their responses, which helps them improve their skills and maintain their motivation, thereby helping them resolve the issue of continued employment.The flow of the identification process in the third embodiment will be described with reference to Fig. 15.

[0378] Step 1:

[0379] The user enters a prompt statement.

[0380] The user inputs a prompt such as a question or instruction into an input field on the terminal. For example, the user might input, "Please tell me the appropriate way to respond to a customer complaint." The input prompt is temporarily stored in the terminal's memory.

[0381] Step 2:

[0382] The terminal sends the prompt to the server.

[0383] The terminal sends the prompt text entered by the user to the server. At this time, the prompt text is sent to the server in an appropriate format (for example, JSON format). The input is the prompt text, and the output is the data sent to the server.

[0384] Step 3:

[0385] The server generates a response using a generative AI model.

[0386] The server inputs the received prompt sentence into a generative AI model (e.g., GPT-4) to generate a response. The input is the prompt sentence, and the output is the generated response. For example, the generated response might be, "First, it's important to listen to the customer's story to the end and show empathy. Then, you need to propose a specific solution and respond quickly."

[0387] Step 4:

[0388] The server-generated response is evaluated using a quality evaluation algorithm.

[0389] The server inputs the generated response into a quality evaluation algorithm (e.g., a BERT-based evaluation model) to evaluate the quality of the response. The input is the generated response, and the output is the evaluation result. For example, a score or comment such as "The quality of the response is high" is output.

[0390] Step 5:

[0391] The server transmits the evaluation results to the terminal.

[0392] The server sends the evaluation result to the terminal together with the response. The input is the evaluation result and the generated response, and the output is the data sent to the terminal.

[0393] Step 6:

[0394] The terminal displays the evaluation results to the user.

[0395] The terminal displays the evaluation results received from the server and the generated response to the user. The input is the evaluation results and the generated response, and the output is the data displayed to the user. The user can refer to the displayed evaluation results and use them to improve their own skills.

[0396] (Application example 3)

[0397] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0398] Conventional complaint handling support systems do not adequately improve the skills of operators or maintain their motivation, leaving the issue of continued employment unresolved. Furthermore, there is a lack of a way to evaluate the quality of responses generated by AI, resulting in inconsistent response quality among operators. Furthermore, the system does not take into consideration the convenience of being an application installed on a smartphone.

[0399] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 3 is realized by the following means. In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further escalation, means for shortening response time, means for evaluating the quality of responses generated by AI, means for supporting the improvement of operators' skills, means for maintaining motivation, and means for supporting the resolution of the issue of continued employment. This enables the improvement of operators' skills and the maintenance of their motivation, thereby resolving the issue of continued employment. Furthermore, by evaluating the quality of responses generated by AI, the quality of responses can be standardized. Furthermore, the convenience of being an application installed on a smartphone is also provided.

[0400] A "contact person" is an employee whose job is to directly respond to customer inquiries and complaints.

[0401] An "operator" is an employee whose job is to deal with customers using communication means such as telephone or chat.

[0402] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[0403] "Means for providing appropriate responses" is a function for generating and providing appropriate responses to customer complaints.

[0404] "Measures to prevent flame wars" is a function to prevent problems from escalating when dealing with complaints.

[0405] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[0406] "Means for evaluating the quality of responses generated by AI" refers to a function for evaluating the quality of responses generated by artificial intelligence.

[0407] "Means to support the improvement of operator skills" refers to a function for improving the response skills of operators.

[0408] "Means for maintaining motivation" refers to a function for maintaining the motivation of operators for their work.

[0409] The "means for supporting the resolution of issues related to continued employment" is a function for resolving issues related to the continued employment of operators.

[0410] The "means for providing a sense of mental security" is a function for providing a sense of mental security to the operator.

[0411] The "means for achieving equalization of response skills" is a function for equalizing the response skills of the agents.

[0412] An "application installed on a smartphone" is software that is installed on a smartphone and used.

[0413] The system for carrying out the present invention is configured as a complaint handling support system. The system includes a server, a terminal (such as a smartphone) used by an operator, and a generative AI model.

[0414] The server includes means for providing appropriate responses to complaints, preventing flame wars, shortening response times, evaluating the quality of responses generated by AI, helping operators improve their skills, maintaining motivation, and helping to resolve issues related to continued employment.

[0415] The device used by the operator includes an application installed on a smartphone that evaluates the quality of the AI-generated responses in real time and provides feedback when the operator handles a claim.

[0416] Specifically, when an operator handles a complaint, the user's input and the AI-generated response are input into the application. The server uses a generative AI model (e.g., GPT-3) to evaluate the quality of the AI-generated response. The evaluation results are displayed as feedback on the operator's device. Based on this feedback, the operator can improve their skills.

[0417] The hardware used includes a server and the operator's smartphone, and the software used includes Python, the OpenAI API, and a smartphone application.

[0418] As a concrete example, the following prompt sentence can be used:

[0419] Example prompt sentence:

[0420] User Input: How do I return an item?

[0421] AI response: To return an item, please log in to your account and select the item you wish to return from your order history. Then, select the reason for return and complete the return process.

[0422] Rate the quality of this response.

[0423] Using this prompt, the server uses a generative AI model to evaluate the quality of the response and provide feedback to the agent, helping them improve their skills, maintain motivation, and solve the challenge of retaining employees.

[0424] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[0425] Step 1:

[0426] When a user handles a complaint, the user's input and the response generated by the AI ​​are entered into the terminal. The input data is the content of the user's complaint and the AI's response text.

[0427] Step 2:

[0428] The terminal sends the user's complaint and the AI's response text to the server. The transmitted data is the user's input text and the AI's response text.

[0429] Step 3:

[0430] The server sends the received user input text and AI response text as a prompt to the generative AI model. An example of a prompt sentence is, "User input: How do I return an item? AI response: To find out how to return an item, first log in to your account and select the item you want to return from your order history. Then, select a reason for return and complete the return process. Please rate the quality of this response."

[0431] Step 4:

[0432] The generative AI model evaluates the quality of the AI's response based on the prompt sentence, and the evaluation result is a feedback text about the quality of the response.

[0433] Step 5:

[0434] The server then sends the evaluation results received from the generative AI model to the device, which then sends the feedback text about the quality of the response.

[0435] Step 6:

[0436] The terminal displays the received evaluation results to the user. The displayed data is a feedback text about the quality of the response. The user can improve their skills based on this feedback.

[0437] Through these processing steps, users can evaluate the quality of responses generated by AI in real time when handling complaints and receive feedback, which helps users improve their skills and maintain their motivation, thereby resolving the issue of continued employment.

[0438] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0439] "Example 1"

[0440] One embodiment of the present invention provides a complaint handling support system that incorporates an emotion engine. When a customer service representative or operator receives a complaint, this system uses an emotion engine that recognizes the user's emotions to analyze the user's emotional state. Based on the analysis results, the system generates an appropriate response to prevent further escalation. Specifically, if the system senses that the user is feeling angry or frustrated, it generates a response that offers an apology or a solution. If the user senses that they are confused, it generates a response that provides a clear explanation or guidance. This allows the customer service representative or operator to quickly respond optimally according to the user's emotional state, thereby shortening response time and improving service quality.

[0441] "Example 2"

[0442] Another embodiment of the present invention provides a complaint handling support system that incorporates an emotion engine. When a customer service representative or operator receives a complaint, this system analyzes the user's emotional state using an emotion engine that recognizes the user's emotions. Based on the analysis results, the system generates an appropriate response to prevent further escalation. Furthermore, this system provides customer service representatives and operators with a sense of mental security and achieves the equalization of response skills. Specifically, the system suggests the optimal response according to the user's emotional state, allowing the customer service representative or operator to respond with confidence. Furthermore, the responses provided by the system eliminate differences in response skills between customer service representatives and operators, maintaining consistency in service quality.

[0443] "Example 3"

[0444] Furthermore, as another embodiment of the present invention, a complaint handling support system incorporating an emotion engine is provided. This system analyzes the emotional state of a customer service representative or operator when they receive a complaint using an emotion engine that recognizes the user's emotions. Based on the analysis results, the system generates an appropriate response to prevent further escalation. Furthermore, this system provides customer service representatives and operators with a sense of mental security and achieves the equalization of response skills. Furthermore, this system supports the resolution of the issue of continued employment. Specifically, the responses and feedback provided by the system help customer service representatives and operators improve their skills and maintain their motivation. This reduces the turnover rate of customer service representatives and operators and promotes continued employment.

[0445] The processing flow of each embodiment will be described below.

[0446] "Example 1"

[0447] Step 1: The counter clerk or operator receives the complaint.

[0448] Step 2: The system uses the emotion engine to analyze the user's emotional state.

[0449] Step 3: Based on the analysis results, the system generates an appropriate response.

[0450] Step 4: A contact person or operator responds to the user using the generated response.

[0451] "Example 2"

[0452] Step 1: The counter clerk or operator receives the complaint.

[0453] Step 2: The system uses the emotion engine to analyze the user's emotional state.

[0454] Step 3: Based on the analysis results, the system generates an appropriate response.

[0455] Step 4: A contact person or operator responds to the user using the generated response.

[0456] Step 5: The customer service representative or operator uses the feedback from the system to improve their own response skills.

[0457] "Example 3"

[0458] Step 1: The counter clerk or operator receives the complaint.

[0459] Step 2: The system uses the emotion engine to analyze the user's emotional state.

[0460] Step 3: Based on the analysis results, the system generates an appropriate response.

[0461] Step 4: A contact person or operator responds to the user using the generated response.

[0462] Step 5: The customer service representative or operator uses the feedback from the system to improve their own response skills.

[0463] Step 6: Tellers and operators use the feedback they receive from the system to maintain their motivation and reduce turnover.

[0464] Example 1

[0465] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0466] When dealing with complaints, it is difficult for the customer service representative or operator to provide an appropriate response quickly. There is also a high risk that the complaint will escalate and become out of hand, and the longer the response time, the lower the customer satisfaction will be. Furthermore, there is a large variation in the response skills of the representatives, which places a heavy mental burden on them. To solve these issues, a system is needed that can accurately analyze the content of the complaint, grasp the user's emotional state, and quickly generate an appropriate response.

[0467] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0468] In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further flaming, means for realizing a reduction in response time, means for analyzing the content of the complaint, means for analyzing the emotional state of the user, means for generating an appropriate response, and means for displaying the generated response, thereby enabling the efficiency and quality of complaint handling to be improved.

[0469] The "Complaint Response Support System" is a system that provides appropriate responses to complaints, prevents them from escalating into flames, and shortens response time.

[0470] The "means for providing an appropriate response" is a function that analyzes the content of the complaint and generates an optimal response based on the user's emotional state.

[0471] "Measures to prevent flame wars" is a function to prevent complaints from escalating and becoming major problems.

[0472] "Means for reducing response time" refers to a function that quickly generates responses to complaints, enabling counter staff or operators to respond quickly.

[0473] The "means for analyzing the content of a claim" is a function that analyzes the text data of a claim using natural language processing technology and understands the content of the claim.

[0474] The "means for analyzing the user's emotional state" is a function for extracting emotions from the user's complaint text and identifying the emotional state, such as anger, dissatisfaction, or confusion.

[0475] The "means for generating an appropriate response" is a function that generates an optimal response according to the emotional state of the user based on the analysis results.

[0476] The "means for displaying the generated response" is a function for sending the generated response from the server to the terminal and displaying it so that the person at the counter or the operator can check it.

[0477] This invention relates to a complaint handling support system for customer service representatives and operators. This system provides appropriate responses to complaints, prevents complaints from escalating, and shortens response times. Specific embodiments of this system are described below.

[0478] Hardware and software used

[0479] server

[0480] The server plays a central role in the complaint handling support system. The server uses the following software and hardware:

[0481] Hardware: General cloud server (e.g., AWS (registered trademark) EC2 instance (registered trademark))

[0482] software:

[0483] Auto-response generation module (e.g., generative AI model)

[0484] Emotion engines (e.g., natural language processing engines)

[0485] Database (e.g. SQL database)

[0486] Terminal

[0487] A terminal is a device used by a teller or operator. A terminal uses the following hardware and software:

[0488] Hardware: General personal computer (e.g., Windows (registered trademark) PC, iPad (registered trademark))

[0489] software:

[0490] Claims Input Interface

[0491] Response Display Interface

[0492] Program processing

[0493] Entering and submitting a claim

[0494] When a user enters a complaint, the terminal sends this information to the server. For example, if a user enters "the product has not arrived," the terminal packages this text data in JSON format and sends it to the server using the HTTPS protocol.

[0495] Claims Analysis

[0496] The server analyzes the text data of the received complaint. Specifically, it uses a generative AI model to perform natural language processing (NLP) to understand the content of the complaint. For example, if a complaint states "the product has not arrived," the server classifies it as a "delivery issue."

[0497] Emotional state analysis

[0498] The server uses an emotion engine to analyze the user's emotional state. It extracts emotions such as anger, frustration, and confusion from the complaint text. For example, for a complaint that "the product has not arrived," the server identifies the emotion "frustrated."

[0499] Generating a response

[0500] The server generates an appropriate response based on the analysis results. Using a generative AI model, it generates a response that reflects the user's emotional state. For example, it generates a response like, "Sorry, we'll check your order. Please wait a moment."

[0501] Sending and Displaying Responses

[0502] The server sends the generated response to the terminal. The terminal displays the response received from the server on its screen. The customer service representative or operator checks this response and takes appropriate action for the user.

[0503] Specific examples

[0504] For example, if a user inputs a complaint stating, "The product has not arrived," the device sends this information to the server. The server then uses a generative AI model to analyze the content of the complaint and an emotion engine to identify the user's emotional state. Based on the analysis results, the server generates a response such as, "We're sorry. We will check your order. Please wait a moment." and sends it to the device. The customer service representative provides this response to the user and attempts to resolve the issue.

[0505] Prompt Sentence Examples

[0506] "If a user enters a complaint such as 'item not received,' how would the server generate a response?"

[0507] In this way, the present invention realizes improved efficiency and quality in handling complaints.

[0508] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0509] Step 1:

[0510] The user enters a claim

[0511] The user inputs a complaint through the interface of the complaint handling support system. For example, the user inputs "The product has not arrived." The input complaint is sent to the terminal as text data. The input data is text that includes the content of the complaint.

[0512] Step 2:

[0513] The device sends the complaint to the server

[0514] The terminal transmits the claims entered by the user to the server in real time. Specifically, the text data of the claims is packaged in JSON format and transmitted to the server using the HTTPS protocol. The input data is the text data of the claims, and the output data is the JSON-formatted data transmitted to the server.

[0515] Step 3:

[0516] The server parses the content of the claim.

[0517] The server analyzes the text data of the received complaint. Specifically, it uses a generative AI model to perform natural language processing (NLP) to understand the content of the complaint. For example, for a complaint that "the product has not arrived," the server classifies it as a "problem with delivery." The input data is the text data of the complaint, and the output data is the classification information of the complaint as a result of the analysis.

[0518] Step 4:

[0519] The server analyzes the user's emotional state

[0520] The server uses an emotion engine to analyze the user's emotional state. Emotions such as anger, frustration, and confusion are extracted from the complaint text. For example, for a complaint that "the product has not arrived," the server identifies the emotion "frustration." The input data is the complaint text data, and the output data is the emotion analysis results.

[0521] Step 5:

[0522] The server generates an appropriate response

[0523] The server generates an appropriate response based on the analysis results. Using a generative AI model, it generates a response that corresponds to the user's emotional state. For example, it generates a response such as, "We're sorry. We will check your order. Please wait a moment." The input data is the complaint classification information and the emotion analysis results, and the output data is the generated response text.

[0524] Step 6:

[0525] The server sends the response to the device

[0526] The server sends the generated response to the terminal. Specifically, it packages the response text data in JSON format and sends it to the terminal using the HTTPS protocol. The input data is the generated response text, and the output data is the JSON-formatted data sent to the terminal.

[0527] Step 7:

[0528] The terminal displays the response

[0529] The terminal displays the response received from the server on the screen. The customer service representative or operator checks this response and takes appropriate action for the user. The input data is the response text sent from the server, and the output data is the response displayed on the screen.

[0530] Step 8:

[0531] The user receives a response

[0532] The user receives a response provided by the agent or operator, which informs the user of the next steps to take toward resolving the problem. The input data is the response provided by the agent or operator, and the output data is the response received by the user.

[0533] (Application example 1)

[0534] Next, a description will be given of Application Example 1 of Embodiment Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0535] Conventional complaint handling systems had the problem of being unable to properly grasp the emotional state of customers and respond quickly and appropriately. Furthermore, they were unable to predict complaint escalation and take measures in advance, which led to a risk of delayed responses and a decline in customer satisfaction. Furthermore, there was also the problem of insufficient management of response history, meaning that past response history could not be utilized to help with future responses.

[0536] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means. In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further escalation, means for shortening response time, means for analyzing the user's emotional state using an emotion engine, means for generating an appropriate response based on the analysis results, means for predicting the possibility of escalation, and means for managing complaint response history. This makes it possible to properly grasp the customer's emotional state and respond quickly and appropriately. Furthermore, by predicting complaint escalation and taking measures in advance, response delays can be prevented and customer satisfaction can be expected to improve. Furthermore, response history management is enhanced, and past response history can be utilized to help with future responses.

[0537] A "customer service representative" is an employee whose role is to directly receive and respond to inquiries and complaints from customers.

[0538] An "operator" is a staff member whose role is to operate a system or machine and carry out a specific task.

[0539] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[0540] "Means for providing an appropriate response" is a function for generating an appropriate response to a complaint and providing it to the customer.

[0541] "Measures to prevent flame wars" are functions to prevent complaints from spreading and problems from becoming more serious.

[0542] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[0543] An "emotion engine" is a technology for analyzing a user's emotional state.

[0544] The "means for analyzing the user's emotional state" is a function for analyzing the user's emotions using an emotion engine.

[0545] The "means for generating an appropriate response" is a function for generating an appropriate response based on the emotion analysis results.

[0546] "Means for predicting the possibility of escalation" is a function for predicting the possibility of a problem becoming serious based on the content of the complaint and the emotional state.

[0547] The "means for managing complaint handling history" is a function for recording and managing past complaint handling history.

[0548] As an embodiment of the present invention, a complaint handling support system for an online shopping site will be described as an example. This system operates in cooperation with three entities: a server, a terminal, and a user.

[0549] System configuration

[0550] 1. Server:

[0551] Emotion engine: A software module for analyzing the user's emotional state. Specifically, it uses a Python emotion analysis library (e.g., TextBlob (registered trademark) or VADER (registered trademark)).

[0552] Generative AI model: An AI model for generating appropriate responses to complaints. Specifically, we use generative AI models such as GPT-3.

[0553] Escalation prediction module: A machine learning model for predicting the likelihood of escalation based on the content of the complaint and the emotional state. Specifically, it uses machine learning libraries such as scikit-learn.

[0554] History management system: A database system for managing the history of complaint responses. Specifically, it uses a database management system such as MySQL (registered trademark) or PostgreSQL (registered trademark).

[0555] 2. Terminal:

[0556] Smartphone app: An application that allows users to input complaints and communicate with the server. This application provides a user interface, allows users to input complaints, displays sentiment analysis results, and receives responses.

[0557] 3. User:

[0558] Customer: Responsible for entering complaints and receiving responses from the system.

[0559] System Operation

[0560] 1. Complaint Reception:

[0561] A user inputs a complaint through a smartphone app, for example, "I'm very unhappy because I didn't receive my product!"

[0562] 2. Emotion analysis:

[0563] The server's emotion engine analyzes the complaint text and determines the user's emotional state. For example, it detects "anger" from the input complaint.

[0564] 3. Auto-response generation:

[0565] The server's generative AI model generates an appropriate response based on the sentiment analysis results, such as "We apologize for the inconvenience. We will investigate this immediately and provide you with a solution."

[0566] 4. Escalation Prediction:

[0567] The server's escalation prediction module predicts the risk of escalation based on the content of the complaint and the emotional state of the customer. For example, it determines the risk as "high risk."

[0568] 5. History Management:

[0569] The server's history management system stores the complaints and their response history in a database, which can be used to handle future complaints.

[0570] Examples and prompts

[0571] Examples:

[0572] Complaint: "I'm very unhappy because I haven't received my item!"

[0573] Sentiment analysis result: Anger

[0574] Automated response: "We apologize for the inconvenience. We will investigate this and provide you with a solution shortly."

[0575] Escalation Risk: High

[0576] Example prompt sentence:

[0577] Customer complaint: "I'm very unhappy because I didn't receive my item!"

[0578] Sentiment analysis result: Anger

[0579] Generate an appropriate response.

[0580] In this way, the server, terminal, and user work together to accurately grasp the emotional state of the customer and respond quickly and appropriately. It also makes it possible to predict complaint escalation and take measures in advance, which prevents delays in response and is expected to improve customer satisfaction. Furthermore, it also improves response history management, making it possible to utilize past response history for future responses.

[0581] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0582] Step 1:

[0583] A user inputs a complaint through a smartphone app. The input complaint text is sent to the server. For example, a complaint such as "I'm very unhappy because my product hasn't arrived!" is input.

[0584] Step 2:

[0585] The emotion engine on the server receives the complaint text and performs emotion analysis. It analyzes the input complaint text and determines the user's emotional state. For example, it uses an emotion analysis library (e.g., TextBlob or VADER) to detect the emotion "anger." The output is the emotional state (e.g., anger).

[0586] Step 3:

[0587] The generative AI model on the server receives the emotion analysis results and generates an appropriate response. Using the emotional state (e.g., anger) and the complaint text as input, a prompt sentence is input to the generative AI model (e.g., GPT-3) to generate a response. For example, a response might be generated that says, "We apologize for the inconvenience. We will investigate this immediately and provide you with a solution." The output is the generated response.

[0588] Step 4:

[0589] The server's escalation prediction module receives the complaint text and emotional state and predicts the risk of escalation. Using the complaint text and emotional state as input, it predicts the escalation risk using a machine learning model (e.g., scikit-learn). For example, it is determined to be "high risk." The output is the escalation risk.

[0590] Step 5:

[0591] The server's history management system stores the complaint and its response history in a database. The complaint text, emotional state, generated responses, and escalation risk are input and stored in a database management system (e.g., MySQL or PostgreSQL). The output is the stored history data.

[0592] Step 6:

[0593] The server sends the generated response to the device. The generated response is sent as input to the user's smartphone app. For example, a response such as "We apologize for the inconvenience. We will investigate this immediately and provide you with a solution" is displayed to the user. The output is the response displayed to the user.

[0594] Step 7:

[0595] The user checks the response through the smartphone app and enters additional complaints or feedback as necessary. If the user checks the response and is satisfied, the complaint handling is completed. If there are additional complaints or feedback, the process starts again from step 1. The output is the user's satisfaction level and additional complaints.

[0596] Example 2

[0597] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0598] In conventional complaint handling systems, it is difficult for the customer service representatives and operators to provide appropriate responses to customer inquiries, placing a heavy mental burden on them. Another issue is that there is a variation in response skills, making it difficult to maintain consistency in service quality. Furthermore, when handling complaints, it is necessary to properly recognize customer emotions and prevent the situation from escalating into a heated argument, but there is a lack of effective means to achieve this.

[0599] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0600] In this invention, the server includes means for generating an appropriate response to an inquiry, means for calculating a confidence score for the generated response, means for displaying the confidence score, means for analyzing the emotional state of the user, means for generating an appropriate response based on the analysis result, means for providing a sense of security, and means for achieving leveling of response skills, thereby enabling tellers and operators to provide consistent, high-quality responses to customers while gaining a sense of security.

[0601] The "means for generating an appropriate response to an inquiry" is a technology for automatically generating an appropriate answer based on the content of an inquiry from a user.

[0602] The "means for calculating a confidence score for a generated response" is a technique for evaluating and quantifying the accuracy and reliability of a generated response.

[0603] A "means for displaying a confidence score" is a technique for visually displaying the calculated confidence score to a user.

[0604] "Means for analyzing the user's emotional state" refers to technology for reading and analyzing emotions from the user's statements and actions.

[0605] "Means for generating an appropriate response based on the analysis results" refers to a technology for generating an optimal response to a user based on the results of emotion analysis.

[0606] "Means for providing psychological security" refers to techniques that enable users to gain a sense of security and trust when using a system.

[0607] "Means for achieving standardization of response skills" refers to technology for reducing the variation in response skills among different users and achieving uniform response quality.

[0608] This invention is a system that enables counter staff and operators to respond appropriately to customer inquiries and complaints. This system uses a generative AI model to generate responses to inquiries and calculates and displays their reliability scores, providing users with a sense of mental security and achieving a leveling of response skills. It also uses an emotion engine to analyze the user's emotional state and generate appropriate responses, improving the quality of complaint handling.

[0609] Hardware and software used

[0610] Server: Provides the computational resources to run generative AI models (e.g., OpenAI's GPT-4).

[0611] Device: The device operated by the user (e.g., PC, tablet, smartphone) displays the response and confidence score sent from the server.

[0612] Emotion engine: Software for analyzing a user's emotional state (e.g., Microsoft® Azure® Cognitive Services®).

[0613] Program processing

[0614] 1. The user receives a customer inquiry. For example, the customer asks, "How long is the warranty on this product?"

[0615] 2. The server uses the generative AI model to generate an appropriate response based on the prompt, "How long is the warranty on this product?"

[0616] 3. The server calculates a confidence score for the generated response. For example, the generated response might be "This product has a one-year warranty" with a confidence score of 90%.

[0617] 4. The server sends the generated response and confidence score to the device.

[0618] 5. The device displays the received response and the confidence score to the user. For example, it might display "Response: This product has a one-year warranty (confidence score: 90%)" on the screen.

[0619] 6. The user checks the displayed response and confidence score and provides the response to the customer with confidence.

[0620] Specific examples

[0621] The user (operator) receives an inquiry from a customer.

[0622] The server uses a generative AI model to generate a response to the query, "How long is the warranty on this product?"

[0623] The server calculates a confidence score for the generated response "This product has a one-year warranty" and sends the confidence score of 90% to the terminal.

[0624] The terminal displays the response and the confidence score to the user.

[0625] The user can view the confidence score and provide a response to the customer with confidence.

[0626] Prompt Sentence Examples

[0627] "Generate an appropriate response to a customer inquiry such as 'How long is the warranty on this product?'"

[0628] In this way, this system utilizes AI technology to provide psychological security and standardize response skills.

[0629] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0630] Step 1:

[0631] A user receives an inquiry from a customer.

[0632] Input: Customer inquiry (e.g., "How long is the warranty on this product?")

[0633] Specific operation: A user (operator) receives an inquiry from a customer by means of telephone, chat, or the like.

[0634] Step 2:

[0635] The server generates a response using a generative AI model.

[0636] Input: The inquiry received from the user (e.g., "How long is the warranty period for this product?")

[0637] Data processing: Input the query content as a prompt into a generative AI model (e.g., OpenAI's GPT-4).

[0638] Output: The generated response (e.g., "This product has a one-year warranty")

[0639] Specific operation: The server inputs the query content into the generative AI model and generates an appropriate response.

[0640] Step 3:

[0641] The server calculates a confidence score for the generated response.

[0642] Input: The generated response (e.g., "This product has a one-year warranty")

[0643] Data Calculation: An internal algorithm is used to calculate the confidence score of the response.

[0644] Output: Confidence score (e.g. 90%)

[0645] Specific behavior: The server evaluates the accuracy and reliability of the generated response and calculates a confidence score.

[0646] Step 4:

[0647] The server sends the response and the confidence score to the device.

[0648] Input: A generated response (e.g., "This product has a one-year warranty") and a confidence score (e.g., 90%)

[0649] Data processing: Responses and confidence scores are combined into a single data packet.

[0650] Output: Data packets sent to the device

[0651] Specific operation: The server assembles the generated response and confidence score into a data packet and sends it to the terminal.

[0652] Step 5:

[0653] The device displays the response and confidence score to the user.

[0654] Input: Data packet sent by the server (response and confidence score)

[0655] Data processing: Parse the data packets and extract the response and confidence score.

[0656] Output: The response and confidence score displayed to the user (e.g., "Response: This product has a one-year warranty (confidence score: 90%)")

[0657] Specific operation: The terminal analyzes the received data packet and displays the response and confidence score on the screen.

[0658] Step 6:

[0659] The user sees the displayed response and confidence score.

[0660] Input: The response displayed on the device and a confidence score (e.g., "Response: This product has a one-year warranty (confidence score: 90%)")

[0661] Specific operation: The user (operator) checks the displayed response and the reliability score and determines the reliability of the response.

[0662] Step 7:

[0663] The user provides a response to the customer.

[0664] Input: A verified response (e.g., "This product has a one-year warranty")

[0665] Specific operation: The user (operator) responds to the customer based on the confirmed response. For example, he / she tells the customer, "The warranty period for this product is one year."

[0666] In this way, by performing specific actions at each step, the system provides psychological security and equalizes response skills.

[0667] (Application example 2)

[0668] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0669] In conventional complaint handling systems, the front desk staff or operators were unable to properly grasp the user's emotional state, which resulted in delayed responses or inappropriate responses. Furthermore, there was no way to verify the reliability of the responses generated by AI, which often caused the operators to feel uneasy. Furthermore, there was a problem of inconsistency in service quality due to variations in response skills.

[0670] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0671] In this invention, the server includes a means for providing an appropriate response to a complaint, a means for preventing further escalation, a means for shortening response time, a means for displaying a reliability score for the response generated by AI, a means for analyzing the user's emotional state using an emotion engine, and a means for generating an appropriate response based on the analysis results. This allows the counter staff or operators to properly grasp the user's emotional state and provide a reliable response quickly. Furthermore, by providing a sense of mental security and achieving a leveling of response skills, consistency in service quality can be maintained.

[0672] A "customer service representative" is an employee whose role is to directly receive and respond to inquiries and complaints from customers.

[0673] An "operator" is a staff member whose role is to operate a system or machine and carry out a specific task.

[0674] A "complaint handling support system" is a system that provides support for appropriate responses to customer complaints.

[0675] "Means of providing appropriate responses" refers to the function of generating and providing appropriate responses to customer inquiries and complaints.

[0676] "Measures to prevent flame wars" is a function that takes measures to prevent customer dissatisfaction and anger from escalating.

[0677] "Means to shorten response times" refers to functions that enable quick responses to complaints and inquiries.

[0678] "Means for displaying a confidence score for AI-generated responses" is a function that quantifies and displays the confidence of answers generated by AI.

[0679] An "emotion engine" is software or hardware for analyzing a user's emotional state.

[0680] "Means for analyzing the user's emotional state" is a function that reads and analyzes emotions from the user's statements and actions.

[0681] "Means for generating an appropriate response based on the analysis results" is a function that generates the optimal answer based on the results of sentiment analysis.

[0682] "Means for providing mental security" is a function that provides support to enable operators to perform their work with peace of mind.

[0683] "Means to achieve standardization of response skills" refers to a function that reduces the variation in response skills between operators and maintains uniform service quality.

[0684] To implement this invention, the following hardware and software are used: A smartphone is required as hardware, and Python, OpenAI API, and the emotion_recognition library are used as software.

[0685] The server first receives a complaint or inquiry from the user as input. Then, it uses the emotion_recognition library to analyze the user's emotional state. Based on the analysis results, it uses the OpenAI API to generate an AI response to the user input. It then obtains a confidence score for the generated response and displays it to the operator.

[0686] For example, if a user inputs a complaint such as "My security system has been malfunctioning recently and I'm having trouble with it. What should I do?", the emotion analysis result will be "anger." Based on this analysis result, the AI ​​will generate a response such as "We apologize for the inconvenience. Could you please try restarting the system?" and display a confidence score of 0.85.

[0687] An example of a prompt sentence is as follows:

[0688] Please enter your complaint or inquiry: My security system has been malfunctioning lately. What should I do?

[0689] In this way, the server can properly grasp the user's emotional state and provide a reliable response quickly. Furthermore, it provides a sense of security and standardizes response skills, thereby maintaining consistency in service quality.

[0690] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0691] Step 1:

[0692] A user inputs a complaint or inquiry using a smartphone. The input text is sent to the server. The input data is the content of the user's complaint or inquiry.

[0693] Step 2:

[0694] The server passes the received user input to the emotion_recognition library to analyze the user's emotional state. The input data is the text of the user's complaint or inquiry, and the output data is the analyzed emotional state (e.g., "anger," "sadness," "joy," etc.).

[0695] Step 3:

[0696] The server generates an AI response to user input using the OpenAI API based on the analyzed emotional state. The input data is the text of the user's complaint or inquiry and the analyzed emotional state, and the output data is the generated AI response text.

[0697] Step 4:

[0698] The server obtains a confidence score for the generated AI response. The input data is the text of the generated AI response, and the output data is the confidence score (e.g., 0.85).

[0699] Step 5:

[0700] The server displays the generated AI response and confidence score on the operator's smartphone. The input data is the text of the generated AI response and the confidence score, and the output data is the information displayed on the operator's smartphone.

[0701] Step 6:

[0702] The operator reviews the displayed AI response and confidence score and takes appropriate action for the user if necessary. The input data is the text of the AI ​​response that the operator reviews and the confidence score, and the output data is the final response that the operator provides to the user.

[0703] In this way, the server can properly grasp the user's emotional state and provide a reliable response quickly. Furthermore, it provides a sense of security and standardizes response skills, thereby maintaining consistency in service quality.

[0704] Example 3

[0705] Next, a description will be given of a third embodiment of the third embodiment. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0706] With conventional complaint handling systems, it was difficult for the front desk staff and operators to respond appropriately to complaints, resulting in inconsistent response quality. Furthermore, it was not possible to properly grasp the user's emotions when handling complaints, making it difficult to prevent complaints from escalating into flames. Furthermore, there was insufficient effort to improve the skills of operators or maintain their motivation, leading to issues with continued employment. To solve these problems, a system was needed that would improve the quality of complaint handling, shorten response times, and help operators improve their skills and maintain their motivation.

[0707] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[0708] In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further flaming, means for realizing a reduction in response time, means for analyzing a user's emotional state using an emotion analysis engine, means for generating an appropriate response using a generative AI model, means for evaluating the quality of the generated response, means for providing feedback to an operator based on the evaluation results, and means for supporting the improvement of the operator's skills. This makes it possible to improve the quality of complaint handling, shorten response time, and support the improvement of the operator's skills and maintenance of motivation.

[0709] A "customer service representative" is an employee whose role is to directly receive and respond to inquiries and complaints from customers.

[0710] An "operator" is a person who operates the system and handles customer service and complaints.

[0711] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[0712] "Means for providing appropriate responses" is a function for generating and providing appropriate responses to customer complaints.

[0713] "Flame prevention measures" are functions that allow for the proper handling of customer dissatisfaction and anger and prevent the problem from escalating.

[0714] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[0715] An "emotion analysis engine" is software or hardware for analyzing a customer's emotional state and providing the results.

[0716] A "generative AI model" is an artificial intelligence model that generates an appropriate response based on a given prompt.

[0717] The "means for assessing the quality of the response" is a function for assessing the appropriateness and effectiveness of the generated response.

[0718] The "means for providing feedback" is a function for providing the operator with advice and points for improvement based on the results of the response quality evaluation.

[0719] "Means to support skill improvement" is a function that provides training and educational programs to improve the response skills of operators.

[0720] "Means for providing a sense of mental security" refers to a function that reduces the stress and anxiety that operators feel while handling complaints, and provides a sense of security.

[0721] "Means to achieve standardization of response skills" refers to a function that reduces the variation in response skills between operators and maintains a consistent level of quality.

[0722] "Means to support the resolution of employment continuity issues" refers to functions to reduce the turnover rate of operators and support long-term employment continuity.

[0723] This invention is a support system for enabling counter staff and operators to appropriately handle customer complaints. The purpose of this system is to improve the quality of complaint handling, shorten response times, and support the improvement of operator skills and maintenance of motivation.

[0724] Hardware and software used

[0725] Hardware: Servers, devices (PCs, tablets, smartphones)

[0726] Software: Generative AI models (e.g., GPT-4), sentiment analysis engines (e.g., Affectiva®), database management systems (e.g., MySQL)

[0727] Specific operation of the system

[0728] Receiving a complaint

[0729] User: A customer service representative or operator receives a complaint from a customer.

[0730] Terminal: The counter staff or operator enters the details of the complaint into the terminal. For example, if a customer complains that the product has not arrived, the details are entered in text format.

[0731] Emotion analysis

[0732] Terminal: Sends the entered complaint details to the server.

[0733] Server: Receives complaint content and sends it to the sentiment analysis engine.

[0734] Sentiment analysis engine: Analyzes the content of complaints and identifies the user's emotional state. For example, it can be analyzed as "anger."

[0735] Server: Receives the analysis results from the sentiment analysis engine and proceeds to the next processing step.

[0736] Response Generation

[0737] Server: Based on the results of the sentiment analysis, the server sends a prompt to the generative AI model. For example, it sends a prompt such as, "The customer is angry because the product has not arrived. Please generate an appropriate response."

[0738] Generative AI model: Generates an appropriate response based on the prompt, for example, "Dear customer, we are very sorry. We will investigate this immediately and provide a solution as soon as possible."

[0739] Server: Receives the generated response and proceeds to the next processing step.

[0740] Response quality assessment

[0741] Server: Runs algorithms to evaluate the quality of the generated responses, e.g., their appropriateness, impact on customer satisfaction, etc.

[0742] Server: Stores the evaluation results in a database and generates feedback.

[0743] Feedback and skill development support

[0744] Server: Generates feedback to the operator based on the evaluation results. For example, "Your response was appropriate, but it would be better if you provided a more specific solution."

[0745] Terminal: Operators receive feedback and undergo training programs to improve their skills, for example, through simulation training to improve the quality of their responses.

[0746] Examples of concrete examples and prompts

[0747] Example: A customer receives a complaint that the product has not arrived.

[0748] Terminal: The counter staff enters the details of the complaint.

[0749] Server: Sends the complaint content to the sentiment analysis engine.

[0750] Sentiment analysis engine: Analyzes the customer's emotional state as "anger."

[0751] Server: Sends the prompt statement "The customer is angry because the product has not arrived" to the generative AI model.

[0752] Generative AI model: Generates the response, "Dear customer, we are very sorry. We will investigate this immediately and provide a solution as soon as possible."

[0753] Server: Evaluates the quality of the response and provides feedback.

[0754] Example prompt sentence:

[0755] "A customer is upset that their item hasn't arrived. Generate an appropriate response."

[0756] "A customer is unhappy with the quality of service. Please generate an appropriate response."

[0757] This system improves the skills of operators and maintains their motivation, thereby resolving the issue of continued employment. The flow of the identification process in the third embodiment will be described with reference to FIG.

[0758] Step 1: Receiving the claim

[0759] User: A customer service representative or operator receives a complaint from a customer. For example, a customer complains that the product has not arrived.

[0760] Terminal: The counter staff or operator enters the details of the claim into the terminal. The entered details of the claim are saved in text format.

[0761] Input: Customer complaint (e.g. "Product not received").

[0762] Output: Claims data in text format.

[0763] Step 2: Sentiment Analysis

[0764] Terminal: Sends the entered complaint details to the server.

[0765] Server: Receives complaint content and sends it to the sentiment analysis engine.

[0766] Sentiment analysis engine: Analyzes the content of complaints and identifies the user's emotional state. For example, it can be analyzed as "anger."

[0767] Input: Claim data in text format.

[0768] Output: Sentiment analysis result (e.g. "Anger").

[0769] Step 3: Response Generation

[0770] Server: Based on the results of the sentiment analysis, the server sends a prompt to the generative AI model. For example, it sends a prompt such as, "The customer is angry because the product has not arrived. Please generate an appropriate response."

[0771] Generative AI model: Generates an appropriate response based on the prompt, for example, "Dear customer, we are very sorry. We will investigate this immediately and provide a solution as soon as possible."

[0772] Input: Sentiment analysis results and prompt sentence.

[0773] Output: The generated response sentence.

[0774] Step 4: Response quality assessment

[0775] Server: Runs algorithms to evaluate the quality of the generated responses, e.g., their appropriateness, impact on customer satisfaction, etc.

[0776] Server: Stores the evaluation results in a database and generates feedback.

[0777] Input: The generated response sentence.

[0778] Output: Response quality evaluation results and feedback.

[0779] Step 5: Feedback and skill development support

[0780] Server: Generates feedback to the operator based on the evaluation results. For example, "Your response was appropriate, but it would be better if you provided a more specific solution."

[0781] Terminal: Operators receive feedback and undergo training programs to improve their skills, for example, through simulation training to improve the quality of their responses.

[0782] Input: Response quality assessment results and feedback.

[0783] Output: Upskilling of operators and implementation of training programs.

[0784] (Application example 3)

[0785] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0786] Conventional complaint handling systems provided insufficient support for counter staff and operators to respond appropriately to complaints, resulting in inconsistent response quality. Furthermore, they lacked the ability to properly analyze user emotions and generate responses based on those emotions, which led to complaints being handled inappropriately, sometimes resulting in outrage. Furthermore, the mental burden on operators was heavy, raising concerns about continued employment. To solve these issues, a system that analyzes user emotions and generates appropriate responses was needed.

[0787] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[0788] In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further flaming, means for shortening response time, means for analyzing a user's emotions, means for generating an appropriate response based on the analysis results, means for evaluating the quality of the generated response, means for feeding back the evaluation results, means for providing a sense of mental security, means for leveling response skills, and means for supporting the resolution of issues related to continued employment. This enables the front desk staff or operators to respond appropriately according to the user's emotions, improves the quality of complaint handling, prevents flaming, reduces mental burden, and promotes continued employment.

[0789] A "contact person" is a staff member whose role is to directly receive and respond to inquiries and complaints from customers and users.

[0790] An "operator" is a staff member whose role is to operate a system or machine and carry out a specific task.

[0791] A "complaint handling support system" is a system that supports appropriate responses to complaints from customers and users.

[0792] "Means for providing appropriate responses" refers to the function for generating and providing appropriate responses to complaints.

[0793] "Measures to prevent flame wars" is a function to prevent complaints from being handled inappropriately and the problem from escalating.

[0794] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[0795] "Means for analyzing user emotions" is a function for analyzing emotions from user comments and actions.

[0796] The "means for generating an appropriate response" is a function for generating an appropriate response based on the analyzed emotion.

[0797] The "means for evaluating the quality of the generated response" is a function for evaluating the quality of the generated response.

[0798] The "means for feeding back evaluation results" is a function for feeding back the quality of the evaluated response to the operator.

[0799] The "means for providing a sense of mental security" is a function for providing a sense of mental security to the operator.

[0800] The "means for achieving equalization of response skills" is a function for equalizing the response skills of the agents.

[0801] "Means to support the resolution of issues related to continued employment" refers to functions that support the improvement of operators' skills and maintenance of motivation, thereby resolving issues related to continued employment.

[0802] To implement this invention, it is necessary to build a complaint handling support system. This system supports counter staff and operators in appropriately handling complaints, and uses the following hardware and software.

[0803] Hardware

[0804] Server: Serves as the core of the complaint handling support system. It processes and analyzes data and generates responses.

[0805] Terminal: A smartphone or computer used by a teller or operator that receives user input and displays responses from the server.

[0806] software

[0807] OpenAI API: Provides the generative AI model used to generate responses.

[0808] EmotionEngine: A library for analyzing user emotions.

[0809] Database: A system for storing claims and response data.

[0810] Processing flow

[0811] 1. Receiving user input

[0812] The user inputs a complaint through the terminal, for example, "The security alarm keeps ringing and I'm having trouble."

[0813] 2. Emotion analysis

[0814] The server uses the Emotion Engine to analyze emotions from user input. For example, it recognizes the emotion "confused" from the user input.

[0815] 3. Response Generation

[0816] The server generates an appropriate response using the OpenAI API based on the analyzed emotion. An example prompt is as follows:

[0817] User Emotion: Confused

[0818] User Input: My security alarm keeps going off and I'm having trouble.

[0819] Generate an appropriate response.

[0820] 4. Response Quality Assessment

[0821] The server evaluates the quality of the generated response, for example by checking whether it contains keywords such as "thank you" or "appreciation."

[0822] 5. Providing Feedback

[0823] The server then provides feedback on the evaluation results to the operator, allowing the operator to improve their own response skills.

[0824] Specific examples

[0825] If a user types "I'm bothered by the security alarm that keeps going off," the server uses the Emotion Engine to analyze the emotion "confused," then uses the OpenAI API to generate a response like this:

[0826] "I see you're having trouble. We'll take care of the security alarm issue right away. Please wait a moment."

[0827] By evaluating the quality of the generated response and feeding back the evaluation results to the operator, the operator's response skills can be improved.

[0828] In this way, the customer service representative or operator can respond appropriately based on the user's emotions, improving the quality of complaint handling, preventing outrage, reducing mental stress, and promoting continued employment.

[0829] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[0830] Step 1:

[0831] The user inputs a complaint through the terminal. For example, the input complaint might be something like, "My security alarm keeps ringing and I'm having trouble." This input data is sent to the server.

[0832] Step 2:

[0833] The server passes the received user input data to the Emotion Engine, which analyzes the user's emotions. The Emotion Engine analyzes the input data and recognizes the emotion, for example, "confusion." The analysis results are returned to the server.

[0834] Step 3:

[0835] The server generates an appropriate response using the OpenAI API based on the analyzed emotion. The server generates a prompt sentence like the following and sends it to the OpenAI API:

[0836] User Emotion: Confused

[0837] User Input: My security alarm keeps going off and I'm having trouble.

[0838] Generate an appropriate response.

[0839] The OpenAI API generates a response based on this prompt and returns it to the server.

[0840] Step 4:

[0841] The server evaluates the quality of the generated response, for example by checking whether the response contains keywords such as "thank you" or "appreciation," and stores the evaluation results locally.

[0842] Step 5:

[0843] The server then feeds back the evaluation results to the operator, who can then display the feedback on their terminal, allowing them to improve their own response skills.

[0844] Step 6:

[0845] The operator responds appropriately to the user based on the feedback provided by the server, for example, "I see you're having trouble. We'll get back to you about your security alarm issue. Please wait a moment."

[0846] In this way, the customer service representative or operator can respond appropriately based on the user's emotions, improving the quality of complaint handling, preventing outrage, reducing mental stress, and promoting continued employment.

[0847] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0848] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0849] Another example of generative AI is Gemini (registered trademark) (Internet search engine). <url: https: gemini.google.com ?hl="ja">) are listed.

[0850] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0851] [Second embodiment]

[0852] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0853] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0854] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0855] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0856] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0857] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0858] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0859] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0860] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0861] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0862] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0863] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0864] "Example 1"

[0865] A first embodiment of the present invention is a complaint handling support system for front desk staff and operators. This system is equipped with an automatic response generation module using AI (artificial intelligence) as a means of providing appropriate responses to complaints. This module analyzes the content of the complaint and generates an appropriate response. It also has a function to predict the escalation of the complaint as a means of preventing further escalation. Furthermore, as a means of shortening response time, an optimization algorithm is used to speed up response generation. "Example 2"

[0866] A second embodiment of the present invention is a system further including a means for providing a sense of mental security and a means for achieving the standardization of response skills. As a means for providing a sense of mental security, the system is provided with a function for displaying a reliability score for the response generated by the AI. This allows the operator to confirm the reliability of the response proposed by the AI ​​and achieve a sense of mental security. Furthermore, as a means for achieving the standardization of response skills, the operator can refer to the response generated by the AI, thereby reducing the variation in response skills between operators.

[0867] "Example 3"

[0868] A third embodiment of the present invention is a system that further includes a means for supporting the resolution of the issue of continued employment. Specifically, the system is equipped with a function for evaluating the quality of responses generated by AI and supporting the improvement of the operator's skills. This helps improve the operator's skills and maintain their motivation, thereby supporting the resolution of the issue of continued employment.

[0869] The processing flow of each embodiment will be described below.

[0870] "Example 1"

[0871] Step 1: The counter clerk or operator receives the complaint.

[0872] Step 2: The system analyzes the content of the claim.

[0873] Step 3: An automatic response generation module using AI (artificial intelligence) generates an appropriate response.

[0874] Step 4: The system predicts claim escalation.

[0875] Step 5: An optimization algorithm is implemented to speed up response generation, reducing response time.

[0876] "Example 2"

[0877] Step 1: The teller or operator reviews the response generated by the AI.

[0878] Step 2: The system displays a confidence score for the AI-generated response.

[0879] Step 3: The teller or operator decides on the final response based on the reliability score.

[0880] Step 4: Counter staff and operators use the responses generated by the AI ​​as a reference to standardize their response skills.

[0881] "Example 3"

[0882] Step 1: The teller / operator uses the AI-generated response.

[0883] Step 2: The system evaluates the quality of the AI-generated response.

[0884] Step 3: The system provides feedback to help operators improve their skills.

[0885] Step 4: Counter staff and operators use the feedback to improve their skills and maintain motivation.

[0886] Example 1

[0887] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0888] Conventional complaint handling systems did not provide sufficient support for the front desk staff or operators to respond appropriately to complaints, which often resulted in variations in the quality and speed of responses. In addition, there was no way to predict the risk of a complaint escalating, and a delayed response was likely to escalate the situation. Furthermore, with the need to reduce response times, it was difficult to generate a prompt and appropriate response. A system that could solve these issues was needed.

[0889] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0890] In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further escalation, means for shortening response time, means for analyzing the content of the complaint, means for generating automatic responses, means for predicting escalation, and means for accelerating response generation. This improves the quality and speed of complaint handling, reduces the risk of escalation, and makes it possible to provide prompt and appropriate responses.

[0891] The "means for providing an appropriate response to a complaint" is a function for generating and providing an appropriate response to a complaint from a user.

[0892] "Measures to prevent further escalation" is a function to prevent complaints from escalating and the problem from expanding.

[0893] The "means for realizing a reduction in response time" is a function for reducing response time by quickly generating and providing responses to complaints.

[0894] The "means for analyzing the content of a complaint" is a function for analyzing the content of a complaint from a user and identifying the type of problem and the emotion behind it.

[0895] The "means for generating an automatic response" is a function for automatically generating an appropriate response based on the analysis results.

[0896] "Means for predicting escalation" is a function for predicting the possibility of a claim escalating and assessing the risk.

[0897] The "means for speeding up response generation" is a function for optimizing the response generation process and providing a response quickly.

[0898] This invention is a support system for enabling counter staff and operators to respond appropriately to complaints. This system includes multiple means for providing appropriate responses to complaints, preventing flaming, and shortening response times.

[0899] System configuration

[0900] Hardware and Software

[0901] Server: Analyzes complaint content, generates automatic responses, predicts escalations, and speeds up response generation.

[0902] Terminal: The device (PC, smartphone, etc.) through which the user enters claims and receives responses from the server.

[0903] Natural language processing (NLP) technology: Used to analyze the content of claims, specifically using Google Cloud Natural Language API and IBM Watson Natural Language Understanding.

[0904] Generative AI model: Used to generate automated responses. Specifically, we use OpenAI's GPT-3.

[0905] Machine learning algorithms: Used to perform escalation predictions, specifically using scikit-learn and TensorFlow.

[0906] Caching techniques and parallel processing are used to speed up response generation.

[0907] System Operation

[0908] 1. Entering a claim

[0909] The user inputs the details of the complaint from the terminal.

[0910] Example: User types "The product is of poor quality."

[0911] 2. Analysis of the complaint

[0912] The server receives the complaint sent by the user.

[0913] The server parses the claim content using the Google Cloud Natural Language API.

[0914] Specifically, the text of the complaint is sent to an API, which performs sentiment analysis and keyword extraction on the text.

[0915] Example: Extract the keyword "quality issue" from the text "poor product quality" and determine the sentiment as "negative."

[0916] 3. Generate Auto-Responses

[0917] The server generates an appropriate response using a generative AI model (GPT-3) based on the analysis results.

[0918] Specifically, the analysis results are input into a generative AI model as a prompt sentence, and a response sentence is generated.

[0919] Example: Enter the prompt text "The user has entered a complaint that the product quality is poor. Please generate an appropriate response." and get the generated response "I'm sorry. Can you tell me more about the quality of the product?"

[0920] 4. Escalation Prediction

[0921] The server uses machine learning algorithms to predict the likelihood of a claim escalation.

[0922] Specifically, we use scikit-learn to build a model that evaluates the risk of escalation based on past claim data and calculate a risk score for current claims.

[0923] Example: A high risk score is assigned to a complaint that has not been resolved despite repeated contact.

[0924] 5. Providing a Response

[0925] The server provides the generated response to the user.

[0926] Specifically, the generated response sentence is sent to the terminal and displayed on the user's screen.

[0927] Example: "Sorry, can you tell me more about the quality of your product?" is the response displayed on the user's device.

[0928] 6. Faster response generation

[0929] The server uses caching techniques and parallel processing to speed up response generation.

[0930] Specifically, past responses to the same type of claim are cached and reused, and processing speed is improved by processing multiple claims in parallel.

[0931] Example: If the response to a complaint about "poor quality of product" is stored in the cache, retrieve the response from the cache and serve it quickly.

[0932] Specific examples

[0933] The user inputs a complaint saying that the product is of poor quality.

[0934] The server sends the complaint content to the Google Cloud Natural Language API and extracts the keyword "quality issue" and the sentiment "negative."

[0935] The server inputs the prompt text "The user has input a complaint that 'the quality of the product is poor.' Please generate an appropriate response." into the generative AI model (GPT-3), which generates the response "I'm sorry. Could you tell me more about the quality of the product?"

[0936] The server calculates the risk score for escalation using scikit-learn and assigns a high risk score.

[0937] The server sends the generated response to the user's terminal and displays it on the user's screen.

[0938] The server caches responses to the same type of claim to speed up subsequent response generation.

[0939] The above is an embodiment of the present invention.

[0940] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0941] Step 1:

[0942] Entering a claim

[0943] The user inputs the details of the complaint from the terminal.

[0944] Input: The complaint entered by the user (e.g., "The product is of poor quality").

[0945] Output: The claim is sent to the server.

[0946] Specific operation: The user enters the details of the complaint into the input form on the terminal and presses the send button.

[0947] Step 2:

[0948] Receiving complaint details

[0949] The server receives the complaint sent by the user.

[0950] Input: The complaint sent by the user.

[0951] Output: The claim is saved on the server.

[0952] Specific operation: The server receives the HTTP request and saves the claim details in the database.

[0953] Step 3:

[0954] Analysis of complaints

[0955] The server parses the claim content using the Google Cloud Natural Language API.

[0956] Input: Saved claim details.

[0957] Output: Analysis results (keywords, sentiment analysis results, etc.).

[0958] Specific operation: The server sends the claim content to the API and obtains the analysis results returned by the API.

[0959] Step 4:

[0960] Generate an auto-response

[0961] The server generates an appropriate response using a generative AI model (GPT-3) based on the analysis results.

[0962] Input: Analysis results.

[0963] Output: The generated response sentence.

[0964] Specific operation: The server inputs the analysis result as a prompt sentence into the generative AI model and obtains the generated response sentence.

[0965] Example: Enter the prompt text "The user has entered a complaint that the product quality is poor. Please generate an appropriate response." to generate the response "I'm sorry. Can you tell me more about the quality of your product?"

[0966] Step 5:

[0967] Escalation Prediction

[0968] The server uses machine learning algorithms to predict the likelihood of a claim escalation.

[0969] Input: Claim details and analysis results.

[0970] Output: Escalation risk score.

[0971] What happens: The server uses a model built from past claims data to calculate a risk score for the current claim.

[0972] Example: A high risk score is assigned to a complaint that has not been resolved despite repeated contact.

[0973] Step 6:

[0974] Providing a response

[0975] The server provides the generated response to the user.

[0976] Input: The generated response sentence.

[0977] Output: The response is displayed on the user's terminal.

[0978] Specific operation: The server sends the generated response to the user's terminal and displays it on the user's screen.

[0979] Example: "Sorry, can you tell me more about the quality of your product?" is the response displayed on the user's device.

[0980] Step 7:

[0981] Faster response generation

[0982] The server uses caching techniques and parallel processing to speed up response generation.

[0983] Input: Past complaints and responses.

[0984] Output: The cached response.

[0985] What it does: The server caches and reuses past responses to the same type of claim, and processes multiple claims in parallel to improve processing speed.

[0986] Example: If the response to a complaint about "poor quality of product" is stored in the cache, retrieve the response from the cache and serve it quickly.

[0987] (Application example 1)

[0988] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0989] With conventional complaint handling systems, customer service representatives and operators had to manually handle complaints, resulting in long response times and lower customer satisfaction. It was also difficult to predict the escalation of complaints, meaning that it was not possible to notify superiors at the appropriate time, increasing the risk of complaints escalating. Furthermore, there was a large variation in response skills and a significant mental burden, which led to issues with continued employment.

[0990] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0991] In this invention, the server includes a means for providing appropriate responses to complaints, a means for preventing further escalation, a means for shortening response time, a means for analyzing the content of the complaint and generating an appropriate response, a means for predicting the escalation of the complaint, a means for automatically notifying a superior, a means for generating responses using a generative AI model, and a means for generating responses using prompt sentences. This enables efficient and rapid complaint handling, improving customer satisfaction and reducing the risk of complaints escalating. It also solves the issue of continued employment by leveling out response skills and providing psychological security.

[0992] A "contact person" is a person in charge of directly responding to inquiries and complaints from customers.

[0993] An "operator" is a person who operates a system or machine and is responsible for carrying out a specific task.

[0994] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[0995] "Means for providing appropriate responses" refers to the function for generating and providing appropriate responses to complaints.

[0996] "Measures to prevent flame wars" are functions to prevent complaints from spreading and problems from becoming more serious.

[0997] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[0998] The "means for analyzing the content of the claim and generating an appropriate response" is a function for analyzing the content of the claim and generating an appropriate response based on that analysis.

[0999] The "means for predicting the escalation of a complaint" is a function for predicting the possibility that a complaint will be escalated to a higher-level person in charge.

[1000] The "means for automatically notifying a superior" is a function for automatically notifying a superior when a complaint escalates.

[1001] "Response generation means using a generative AI model" is a function for generating responses to claims using an artificial intelligence model.

[1002] The "means for generating a response using a prompt sentence" is a function for generating a response to a complaint using a specific input sentence (prompt sentence).

[1003] As an example of an embodiment of this invention, we will explain a customer support application for an online shopping site. This system uses AI to automatically generate appropriate responses to customer complaints and inquiries, enabling quick responses. It also has a function to predict complaint escalation and automatically notify higher-level personnel as necessary.

[1004] System Configuration

[1005] The system consists of the following major components:

[1006] 1. Complaint Analysis Module: Analyzes the content of the complaint and generates an appropriate response.

[1007] 2. Escalation prediction module: predicts the escalation of claims.

[1008] 3. Notification module: Automatically notify superiors when escalation is predicted.

[1009] 4. Generative AI model: An artificial intelligence model for generating responses to claims.

[1010] 5. Prompt generation module: Generates a response using a specific input sentence (prompt sentence).

[1011] Hardware and Software

[1012] The system uses an internet-enabled smartphone or PC as hardware, and Python and OpenAI API as software.

[1013] Processing flow

[1014] 1. Entering the details of the complaint: The user enters the details of the complaint.

[1015] 2. Response generation: The claim analysis module analyzes the content of the claim and generates an appropriate response using a generative AI model.

[1016] 3. Escalation Prediction: The escalation prediction module predicts the likelihood of a claim escalation.

[1017] 4. Notification: When an escalation is predicted, the notification module automatically notifies superiors.

[1018] Specific examples

[1019] For example, if a customer enters a complaint stating, "I did not receive the product, so I would like a refund," the system operates as follows:

[1020] 1. Entering the complaint details: The user enters, "I did not receive the product, so I would like a refund."

[1021] 2. Response generation: The complaint analysis module analyzes this content and uses a generative AI model to generate a response such as, "We apologize for the inconvenience. We will immediately investigate the issue of the non-delivery of the product and proceed with the refund process."

[1022] 3. Escalation Prediction: The escalation prediction module predicts that the likelihood of escalation for this claim is "high."

[1023] 4. Notification: The notification module automatically notifies the superior, who then takes over the response.

[1024] Prompt Sentence Examples

[1025] Please enter your complaint: I did not receive my item and would like a refund.

[1026] In this way, the efficiency of customer support on the online shopping site can be improved, and customer satisfaction can be improved.

[1027] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1028] Step 1:

[1029] The user inputs the complaint details.

[1030] Input: The user enters, "I never received my item and would like a refund."

[1031] Specific operation: The user enters the content of the complaint in text format into an input form on a smartphone or PC.

[1032] Step 2:

[1033] The server receives the complaint content and sends it to the complaint analysis module.

[1034] Input: The claim entered by the user.

[1035] Output: The claim content sent to the claim analysis module.

[1036] Specific operation: The server receives input from the user and sends the content to the complaint analysis module.

[1037] Step 3:

[1038] The claim analysis module analyzes the content of the claim and generates an appropriate response using a generative AI model.

[1039] Input: The claim content sent to the claim analysis module.

[1040] Output: The generated response text.

[1041] Specific operation: The claim analysis module analyzes the content of the claim using natural language processing technology and sends it as a prompt to the generative AI model. The generative AI model generates an appropriate response and returns it to the claim analysis module.

[1042] Step 4:

[1043] The server generates a response and sends it to the user.

[1044] Input: The generated response text.

[1045] Output: The response text that is displayed to the user.

[1046] Specific operation: The server receives the generated response and sends it to the user's device, which displays the response.

[1047] Step 5:

[1048] An escalation prediction module predicts the likelihood of a claim escalation.

[1049] Input: Claim details.

[1050] Output: Escalation prediction (high / low).

[1051] Specific operation: The escalation prediction module analyzes the content of the complaint and predicts the possibility of escalation using a generative AI model. The prediction result is returned to the server.

[1052] Step 6:

[1053] The server receives the predicted results of the escalation and automatically notifies the superior as necessary.

[1054] Input: The predicted outcome of the escalation.

[1055] Output: Notification to superiors.

[1056] Specific operation: The server receives the escalation prediction result, and if the prediction result is "high," it sends a notification to the superior via email or a dedicated application.

[1057] In this way, the system can make complaint handling more efficient and faster, improving customer satisfaction and reducing the risk of complaints escalating.

[1058] Example 2

[1059] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1060] Conventional complaint handling support systems do not provide sufficient support for operators to respond appropriately, and lack means to provide psychological security or to standardize response skills. They also lack effective means to shorten response times or prevent complaints from escalating. This increases the burden on operators and leads to inconsistent response quality, which is an issue.

[1061] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1062] In this invention, the server includes means for receiving an inquiry, means for generating a response using a generative AI model, means for calculating a reliability score for the generated response, means for displaying the response and the reliability score, means for providing an appropriate response to the complaint, means for preventing further flaming, and means for realizing a reduction in response time. This allows the operator to confirm the reliability of the response generated by the AI ​​and gain psychological peace of mind, while also reducing the variation in response skills, shortening response time, and preventing flaming.

[1063] The "means for receiving an inquiry" is a function that enables the server to receive an inquiry from a user.

[1064] "Means for generating a response using a generative AI model" is a function for generating an appropriate response to a received inquiry using a generative AI model.

[1065] The "means for calculating a confidence score for a generated response" is a function for evaluating the reliability of a generated response and calculating a confidence score.

[1066] The "means for displaying the response and the confidence score" is a function for visually displaying the generated response and its confidence score to the user.

[1067] "Means for providing appropriate responses to complaints" refers to a function for providing responses that allow appropriate responses to be made to complaints.

[1068] "Measures to prevent further escalation" is a function that provides preventative measures to prevent problems from escalating when dealing with complaints.

[1069] "Means to shorten response times" refers to functions that improve efficiency in order to shorten the time it takes to respond to complaints.

[1070] "Means for providing psychological reassurance" is a function that allows operators to gain psychological reassurance by confirming the reliability of the AI's responses.

[1071] "Means for achieving standardization of response skills" refers to a function for reducing the variation in response skills among operators and achieving uniform response.

[1072] The "means for supporting the resolution of issues regarding employment continuity" is a function for providing support for the resolution of issues regarding the employment continuity of operators.

[1073] This invention is a complaint handling support system that uses a generative AI model to generate responses and displays their reliability scores, thereby providing peace of mind to operators and achieving a leveling of response skills. Specific embodiments of this system are described below.

[1074] Server Processing

[1075] The server receives a query from the user. The query is a prompt sentence, such as "Please tell me the progress of the new project." The server inputs this query into a generative AI model (e.g., GPT-4) and generates an appropriate response. For the generated response, the server calculates a confidence score. This confidence score is calculated using the AI ​​model's internal evaluation mechanism or an additional evaluation algorithm. The server sends the generated response and confidence score to the terminal.

[1076] Terminal handling

[1077] The device receives the response and confidence score sent from the server. The received response and confidence score are displayed on the device screen. For example, it may say, "Current progress is going well. The next milestone is next week. (Confidence score: 85%)." The device makes it easy to understand visually by displaying a high confidence score in green and a low confidence score in red.

[1078] User Action

[1079] The user checks the response and the confidence score displayed on the device. The confidence score is used to evaluate the reliability of the response. For example, if the confidence score is 85%, the user can trust the response and adopt it. If the confidence score is low, the user can make a second inquiry. When making a second inquiry, the user can add more specific questions by referring to the previous response.

[1080] Specific examples

[1081] Example 1: Means of providing psychological security

[1082] The user sends a query such as "How is the progress of the new project going?" The server uses a generative AI model (e.g., GPT-4) to generate a response such as "Current progress is going well. The next milestone is next week," and calculates a confidence score of 85%. The device displays this response and the confidence score, and the user can check the confidence score and feel reassured.

[1083] Example 2: A means to achieve leveling of response skills

[1084] Operator A receives an inquiry saying, "What are the steps to take in handling a complaint?" The server uses a generative AI model to generate a response saying, "First, listen carefully to the customer, then confirm the details of the problem, and finally propose a solution." Operator A responds using this response as a reference, and Operator B can also refer to the same steps when handling similar inquiries, thereby reducing the variability in response skills.

[1085] Prompt Sentence Examples

[1086] "Please tell me how your new project is progressing."

[1087] "Please tell me the procedure for handling complaints."

[1088] These prompts can be fed into a generative AI model to generate an appropriate response and confidence score.

[1089] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1090] Step 1:

[1091] The server receives a query from a user. When the user inputs a prompt such as "Please tell me the progress of the new project" through the terminal and presses the send button, the server receives the query. The input is the user's query, and the output is the query data stored on the server.

[1092] Step 2:

[1093] The server inputs the received query into the generative AI model. Specifically, the server sends the query content to the generative AI model (e.g., GPT-4) via an API. The input is the query data, and the output is the prompt sent to the generative AI model.

[1094] Step 3:

[1095] The generative AI model generates a response based on the prompt sentence. The generative AI model uses an internal natural language processing algorithm to generate an appropriate response. The input is the prompt sentence, and the output is the generated response text.

[1096] Step 4:

[1097] The server calculates a confidence score for the generated response. The server evaluates the reliability of the response using the AI ​​model's internal evaluation mechanism or additional evaluation algorithms, and calculates the confidence score. The input is the generated response text, and the output is the confidence score.

[1098] Step 5:

[1099] The server sends the generated response and confidence score to the terminal. The server combines the response text and confidence score into a single data packet and sends it to the terminal. The input is the response text and confidence score, and the output is the data packet sent to the terminal.

[1100] Step 6:

[1101] The terminal receives the response and confidence score sent from the server. The terminal analyzes the received data packet and extracts the response text and confidence score. The input is the data packet sent from the server, and the output is the analyzed response text and confidence score.

[1102] Step 7:

[1103] The terminal displays the received response and confidence score to the user. The terminal screen displays "Current progress is good. The next milestone is next week. (Confidence score: 85%)." The input is the parsed response text and confidence score, and the output is the information displayed to the user.

[1104] Step 8:

[1105] The user checks the response and the confidence score displayed on the terminal. The user evaluates the reliability of the response based on the confidence score. The input is the displayed response text and the confidence score, and the output is the user's evaluation result.

[1106] Step 9:

[1107] The user can either adopt the response or ask again. If the confidence score is high, the user can trust the response and adopt it. If the confidence score is low, the user can ask again with more specific questions. The input is the user's evaluation result, and the output is the adoption of a new query or response.

[1108] (Application example 2)

[1109] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1110] In conventional complaint response support systems, operators had to provide appropriate responses to complaints, which placed a heavy mental burden on them, and their response skills often varied. Furthermore, when security operators received AI suggestions, they lacked the means to verify their reliability, making it difficult to respond quickly and appropriately. This resulted in issues such as longer response times and an increased risk of outrage.

[1111] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1112] In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further flaming, means for shortening response time, means for displaying a reliability score for the response generated by the AI, means for reducing the variability in response skills by having an operator refer to the response generated by the AI, and means for a security operator to receive the AI's suggestions and respond while checking the reliability score. This allows the operator to respond quickly and appropriately while gaining peace of mind.

[1113] A "contact person" is a person whose job is to directly respond to inquiries and complaints from customers and users.

[1114] An "operator" is a person whose job is to operate a system or machine and perform a specific task.

[1115] A "complaint handling support system" is a system that supports appropriate responses to complaints from customers and users.

[1116] "Means for providing appropriate responses" refers to the function of generating accurate and effective responses to complaints and providing them to operators.

[1117] "Measures to prevent flame wars" is a function that takes measures to prevent problems from escalating when dealing with complaints.

[1118] "Means for shortening response time" refers to a function for reducing the time it takes to respond to complaints.

[1119] "Means for displaying a reliability score for responses generated by AI" is a function that quantifies the reliability of responses generated by AI and displays it to the operator.

[1120] "Means to reduce variability in response skills by having operators refer to responses generated by AI" is a function that reduces the differences in response skills between operators by having operators refer to responses generated by AI.

[1121] A "security operator" is a person in charge of security-related tasks and whose job is to ensure the safety of systems and facilities.

[1122] A "confidence score" is a number used to evaluate the reliability of a response generated by AI.

[1123] As an embodiment of the present invention, a complaint handling support system used by a security operator will be specifically described.

[1124] System configuration

[1125] The system consists of the following major components:

[1126] 1. Server: Hosts the AI ​​model and generates responses to queries.

[1127] 2. Terminal: A device such as a smartphone or tablet used by a security operator.

[1128] 3. User interface: An application where operators can enter queries and see the AI's responses and confidence scores.

[1129] Hardware and software used

[1130] Hardware: smartphones, tablets, servers

[1131] Software: Flask (a Python web framework), requests (a library for sending HTTP requests), generative AI models

[1132] Data processing and calculation

[1133] The server receives queries from security operators and sends prompts to the generative AI model, which generates responses to the queries and returns the responses to the server, which then calculates a confidence score for the responses and sends it to the operator's terminal.

[1134] Processing flow

[1135] 1. Entering a query: The security operator enters a query through the user interface of the terminal.

[1136] 2. Sending a query: The terminal sends a query to the server.

[1137] 3. Generate response: The server sends a query to the generative AI model and generates a response.

[1138] 4. Calculate confidence score: The server calculates a confidence score for the generated response.

[1139] 5. Displaying the response and score: The server sends the response and confidence score to the terminal for the operator to review.

[1140] Specific examples

[1141] For example, a security operator enters a query such as "What should we do if a suspicious person breaks into a building?" This query is sent to the server, and the generative AI model generates an appropriate response. The server calculates a confidence score for the generated response and displays the response and score on the operator's device.

[1142] Prompt Sentence Examples

[1143] What should you do if a suspicious person breaks into your building?

[1144] In this way, security operators can take prompt and appropriate action while referring to AI suggestions, which will provide peace of mind and reduce the variability in response skills.

[1145] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1146] Step 1:

[1147] A security operator enters a query through a terminal user interface.

[1148] Input: The query entered by the security operator (e.g., "What should we do if a suspicious person breaks into our building?")

[1149] Output: The query is entered into the terminal.

[1150] How it works: The security operator opens the application on their smartphone or tablet, enters a query in the text box, and presses the "send" button.

[1151] Step 2:

[1152] The device sends a query to the server.

[1153] Input: The query typed into the terminal

[1154] Output: A query is sent to the server.

[1155] Specific operation: The terminal application sends the entered query to the server as an HTTP request.

[1156] Step 3:

[1157] The server sends queries to the generative AI model and generates a response.

[1158] Input: The query sent to the server

[1159] Output: The response from the generative AI model

[1160] Specific operation: The server sends the received query as a prompt to the generative AI model, and the generative AI model generates a response to the query and returns it to the server.

[1161] Step 4:

[1162] The server calculates a confidence score for the generated response.

[1163] Input: Response from a generative AI model

[1164] Output: Confidence score

[1165] Specific operation: The server analyzes the content of the generated response and applies a confidence score calculation algorithm to calculate a confidence score.

[1166] Step 5:

[1167] The server sends the response and the confidence score to the device.

[1168] Input: Response and confidence score from the generative AI model

[1169] Output: Response and confidence score sent to the device

[1170] Specific operation: The server combines the response and the confidence score and sends it to the terminal as an HTTP response.

[1171] Step 6:

[1172] The terminal displays the response and confidence score to the security operator.

[1173] Input: The response and confidence score sent by the server

[1174] Output: Response and confidence score displayed to the security operator

[1175] Specific operation: The terminal application displays the received response and confidence score on the screen for the security operator to review.

[1176] Example 3

[1177] Next, a description will be given of Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1178] Conventional complaint handling support systems did not provide enough support for counter staff and operators to respond appropriately, making it difficult to reduce response times, provide psychological security, and standardize response skills.In addition, it was difficult to improve the skills of operators and maintain their motivation, which led to issues with continuing employment.

[1179] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[1180] In this invention, the server includes a means for generating a response to a prompt using a generative AI model, a means for evaluating the quality of the generated response, and a means for providing feedback based on the evaluation results. This allows operators to respond appropriately and quickly, improving their skills and maintaining their motivation. It also provides a sense of security and standardizes response skills, resolving the issue of continued employment.

[1181] A "customer service representative" is an employee whose role is to directly respond to customer inquiries and complaints.

[1182] An "operator" is a person who operates the system and handles customer service and complaints.

[1183] A "complaint handling support system" is a system that provides support for appropriate responses to customer complaints.

[1184] A "generative AI model" is a model that uses artificial intelligence to generate a response to an input prompt.

[1185] A "prompt sentence" is text such as a question or instruction that a user enters into a system.

[1186] A "response" is a text answer or suggestion that a generative AI model generates based on a prompt.

[1187] A "quality assessment algorithm" is an algorithm for assessing the quality of a generated response.

[1188] "Feedback" refers to information provided to operators based on the evaluation results, encouraging them to improve their skills and improve.

[1189] "Skill improvement" refers to an operator improving their own abilities and techniques.

[1190] "Maintaining motivation" refers to operators maintaining their enthusiasm and motivation for their work.

[1191] "Mental security" refers to the sense of peace and security that an operator feels when performing their work.

[1192] "Leveling response skills" refers to ensuring that multiple operators have the same level of response skills.

[1193] "Continued employment" refers to an operator continuing to work at the same workplace for an extended period of time.

[1194] This invention is a support system for helping service representatives and operators respond appropriately to customer complaints. The system uses a generative AI model to generate responses to prompts, evaluates the quality of the responses, and provides feedback on the evaluation results to improve the skills of the operators and maintain their motivation.

[1195] System configuration

[1196] Hardware

[1197] Server: Provides the computational resources to run the generative AI model and quality assessment algorithm.

[1198] Terminal: A device through which a user can enter prompts and view responses and evaluation results from the server.

[1199] software

[1200] Generative AI models: For example, using natural language generation models such as GPT-4.

[1201] Quality evaluation algorithm: For example, use a BERT-based evaluation model.

[1202] Program processing

[1203] The server receives the prompt sentence entered by the user from the device. It then generates a response to the prompt sentence using a generative AI model. The generated response is evaluated by a quality evaluation algorithm, and the evaluation result is sent to the device as feedback. The device then displays the evaluation result and the generated response to the user.

[1204] Specific examples

[1205] For example, if a user enters the prompt "What is the appropriate way to respond to a customer complaint?", the server will process it as follows:

[1206] 1. The server receives the prompt: "Please tell me the appropriate way to respond to a customer complaint."

[1207] 2. Use a generative AI model (e.g., GPT-4) to generate responses like this: "First, listen to the customer fully and demonstrate empathy. Then, offer a concrete solution and respond quickly."

[1208] 3. The server evaluates this response using a quality evaluation algorithm (e.g., a BERT-based evaluation model) and determines, for example, that "the quality of the response is high."

[1209] 4. The evaluation results are sent to the operator as feedback to help improve their skills.

[1210] Other examples of prompts include:

[1211] Please briefly explain the features of your new product.

[1212] "What is the proper way to respond to customer complaints?"

[1213] "Suggest ways to motivate your team."

[1214] This system allows operators to objectively evaluate the quality of their responses, which helps them improve their skills and maintain their motivation, thereby helping them resolve the issue of continued employment.The flow of the identification process in the third embodiment will be described with reference to Fig. 15.

[1215] Step 1:

[1216] The user enters a prompt statement.

[1217] The user inputs a prompt such as a question or instruction into an input field on the terminal. For example, the user might input, "Please tell me the appropriate way to respond to a customer complaint." The input prompt is temporarily stored in the terminal's memory.

[1218] Step 2:

[1219] The terminal sends the prompt to the server.

[1220] The terminal sends the prompt text entered by the user to the server. At this time, the prompt text is sent to the server in an appropriate format (for example, JSON format). The input is the prompt text, and the output is the data sent to the server.

[1221] Step 3:

[1222] The server generates a response using a generative AI model.

[1223] The server inputs the received prompt sentence into a generative AI model (e.g., GPT-4) to generate a response. The input is the prompt sentence, and the output is the generated response. For example, the generated response might be, "First, it's important to listen to the customer's story to the end and show empathy. Then, you need to propose a specific solution and respond quickly."

[1224] Step 4:

[1225] The server-generated response is evaluated using a quality evaluation algorithm.

[1226] The server inputs the generated response into a quality evaluation algorithm (e.g., a BERT-based evaluation model) to evaluate the quality of the response. The input is the generated response, and the output is the evaluation result. For example, a score or comment such as "The quality of the response is high" is output.

[1227] Step 5:

[1228] The server transmits the evaluation results to the terminal.

[1229] The server sends the evaluation result to the terminal together with the response. The input is the evaluation result and the generated response, and the output is the data sent to the terminal.

[1230] Step 6:

[1231] The terminal displays the evaluation results to the user.

[1232] The terminal displays the evaluation results received from the server and the generated response to the user. The input is the evaluation results and the generated response, and the output is the data displayed to the user. The user can refer to the displayed evaluation results and use them to improve their own skills.

[1233] (Application example 3)

[1234] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1235] Conventional complaint handling support systems do not adequately improve the skills of operators or maintain their motivation, leaving the issue of continued employment unresolved. Furthermore, there is a lack of a way to evaluate the quality of responses generated by AI, resulting in inconsistent response quality among operators. Furthermore, the system does not take into consideration the convenience of being an application installed on a smartphone.

[1236] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 3 is realized by the following means. In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further escalation, means for shortening response time, means for evaluating the quality of responses generated by AI, means for supporting the improvement of operators' skills, means for maintaining motivation, and means for supporting the resolution of the issue of continued employment. This enables the improvement of operators' skills and the maintenance of their motivation, thereby resolving the issue of continued employment. Furthermore, by evaluating the quality of responses generated by AI, the quality of responses can be standardized. Furthermore, the convenience of being an application installed on a smartphone is also provided.

[1237] A "contact person" is an employee whose job is to directly respond to customer inquiries and complaints.

[1238] An "operator" is an employee whose job is to deal with customers using communication means such as telephone or chat.

[1239] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[1240] "Means for providing appropriate responses" is a function for generating and providing appropriate responses to customer complaints.

[1241] "Measures to prevent flame wars" is a function to prevent problems from escalating when dealing with complaints.

[1242] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[1243] "Means for evaluating the quality of responses generated by AI" refers to a function for evaluating the quality of responses generated by artificial intelligence.

[1244] "Means to support the improvement of operator skills" refers to a function for improving the response skills of operators.

[1245] "Means for maintaining motivation" refers to a function for maintaining the motivation of operators for their work.

[1246] The "means for supporting the resolution of issues related to continued employment" is a function for resolving issues related to the continued employment of operators.

[1247] The "means for providing a sense of mental security" is a function for providing a sense of mental security to the operator.

[1248] The "means for achieving equalization of response skills" is a function for equalizing the response skills of the agents.

[1249] An "application installed on a smartphone" is software that is installed on a smartphone and used.

[1250] The system for carrying out the present invention is configured as a complaint handling support system. The system includes a server, a terminal (such as a smartphone) used by an operator, and a generative AI model.

[1251] The server includes means for providing appropriate responses to complaints, preventing flame wars, shortening response times, evaluating the quality of responses generated by AI, helping operators improve their skills, maintaining motivation, and helping to resolve issues related to continued employment.

[1252] The device used by the operator includes an application installed on a smartphone that evaluates the quality of the AI-generated responses in real time and provides feedback when the operator handles a claim.

[1253] Specifically, when an operator handles a complaint, the user's input and the AI-generated response are input into the application. The server uses a generative AI model (e.g., GPT-3) to evaluate the quality of the AI-generated response. The evaluation results are displayed as feedback on the operator's device. Based on this feedback, the operator can improve their skills.

[1254] The hardware used includes a server and the operator's smartphone, and the software used includes Python, the OpenAI API, and a smartphone application.

[1255] As a concrete example, the following prompt sentence can be used:

[1256] Example prompt sentence:

[1257] User Input: How do I return an item?

[1258] AI response: To return an item, please log in to your account and select the item you wish to return from your order history. Then, select the reason for return and complete the return process.

[1259] Rate the quality of this response.

[1260] Using this prompt, the server uses a generative AI model to evaluate the quality of the response and provide feedback to the agent, helping them improve their skills, maintain motivation, and solve the challenge of retaining employees.

[1261] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[1262] Step 1:

[1263] When a user handles a complaint, the user's input and the response generated by the AI ​​are entered into the terminal. The input data is the content of the user's complaint and the AI's response text.

[1264] Step 2:

[1265] The terminal sends the user's complaint and the AI's response text to the server. The transmitted data is the user's input text and the AI's response text.

[1266] Step 3:

[1267] The server sends the received user input text and AI response text as a prompt to the generative AI model. An example of a prompt sentence is, "User input: How do I return an item? AI response: To find out how to return an item, first log in to your account and select the item you want to return from your order history. Then, select a reason for return and complete the return process. Please rate the quality of this response."

[1268] Step 4:

[1269] The generative AI model evaluates the quality of the AI's response based on the prompt sentence, and the evaluation result is a feedback text about the quality of the response.

[1270] Step 5:

[1271] The server then sends the evaluation results received from the generative AI model to the device, which then sends the feedback text about the quality of the response.

[1272] Step 6:

[1273] The terminal displays the received evaluation results to the user. The displayed data is a feedback text about the quality of the response. The user can improve their skills based on this feedback.

[1274] Through these processing steps, users can evaluate the quality of responses generated by AI in real time when handling complaints and receive feedback, which helps users improve their skills and maintain their motivation, thereby resolving the issue of continued employment.

[1275] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1276] "Example 1"

[1277] One embodiment of the present invention provides a complaint handling support system that incorporates an emotion engine. When a customer service representative or operator receives a complaint, this system uses an emotion engine that recognizes the user's emotions to analyze the user's emotional state. Based on the analysis results, the system generates an appropriate response to prevent further escalation. Specifically, if the system senses that the user is feeling angry or frustrated, it generates a response that offers an apology or a solution. If the user senses that they are confused, it generates a response that provides a clear explanation or guidance. This allows the customer service representative or operator to quickly respond optimally according to the user's emotional state, thereby shortening response time and improving service quality.

[1278] "Example 2"

[1279] Another embodiment of the present invention provides a complaint handling support system that incorporates an emotion engine. When a customer service representative or operator receives a complaint, this system analyzes the user's emotional state using an emotion engine that recognizes the user's emotions. Based on the analysis results, the system generates an appropriate response to prevent further escalation. Furthermore, this system provides customer service representatives and operators with a sense of mental security and achieves the equalization of response skills. Specifically, the system suggests the optimal response according to the user's emotional state, allowing the customer service representative or operator to respond with confidence. Furthermore, the responses provided by the system eliminate differences in response skills between customer service representatives and operators, maintaining consistency in service quality.

[1280] "Example 3"

[1281] Furthermore, as another embodiment of the present invention, a complaint handling support system incorporating an emotion engine is provided. This system analyzes the emotional state of a customer service representative or operator when they receive a complaint using an emotion engine that recognizes the user's emotions. Based on the analysis results, the system generates an appropriate response to prevent further escalation. Furthermore, this system provides customer service representatives and operators with a sense of mental security and achieves the equalization of response skills. Furthermore, this system supports the resolution of the issue of continued employment. Specifically, the responses and feedback provided by the system help customer service representatives and operators improve their skills and maintain their motivation. This reduces the turnover rate of customer service representatives and operators and promotes continued employment.

[1282] The processing flow of each embodiment will be described below.

[1283] "Example 1"

[1284] Step 1: The counter clerk or operator receives the complaint.

[1285] Step 2: The system uses the emotion engine to analyze the user's emotional state.

[1286] Step 3: Based on the analysis results, the system generates an appropriate response.

[1287] Step 4: A contact person or operator responds to the user using the generated response.

[1288] "Example 2"

[1289] Step 1: The counter clerk or operator receives the complaint.

[1290] Step 2: The system uses the emotion engine to analyze the user's emotional state.

[1291] Step 3: Based on the analysis results, the system generates an appropriate response.

[1292] Step 4: A contact person or operator responds to the user using the generated response.

[1293] Step 5: The customer service representative or operator uses the feedback from the system to improve their own response skills.

[1294] "Example 3"

[1295] Step 1: The counter clerk or operator receives the complaint.

[1296] Step 2: The system uses the emotion engine to analyze the user's emotional state.

[1297] Step 3: Based on the analysis results, the system generates an appropriate response.

[1298] Step 4: A contact person or operator responds to the user using the generated response.

[1299] Step 5: The customer service representative or operator uses the feedback from the system to improve their own response skills.

[1300] Step 6: Tellers and operators use the feedback they receive from the system to maintain their motivation and reduce turnover.

[1301] Example 1

[1302] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1303] When dealing with complaints, it is difficult for the customer service representative or operator to provide an appropriate response quickly. There is also a high risk that the complaint will escalate and become out of hand, and the longer the response time, the lower the customer satisfaction will be. Furthermore, there is a large variation in the response skills of the representatives, which places a heavy mental burden on them. To solve these issues, a system is needed that can accurately analyze the content of the complaint, grasp the user's emotional state, and quickly generate an appropriate response.

[1304] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1305] In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further flaming, means for realizing a reduction in response time, means for analyzing the content of the complaint, means for analyzing the emotional state of the user, means for generating an appropriate response, and means for displaying the generated response, thereby enabling the efficiency and quality of complaint handling to be improved.

[1306] The "Complaint Response Support System" is a system that provides appropriate responses to complaints, prevents them from escalating into flames, and shortens response time.

[1307] The "means for providing an appropriate response" is a function that analyzes the content of the complaint and generates an optimal response based on the user's emotional state.

[1308] "Measures to prevent flame wars" is a function to prevent complaints from escalating and becoming major problems.

[1309] "Means for reducing response time" refers to a function that quickly generates responses to complaints, enabling counter staff or operators to respond quickly.

[1310] The "means for analyzing the content of a claim" is a function that analyzes the text data of a claim using natural language processing technology and understands the content of the claim.

[1311] The "means for analyzing the user's emotional state" is a function for extracting emotions from the user's complaint text and identifying the emotional state, such as anger, dissatisfaction, or confusion.

[1312] The "means for generating an appropriate response" is a function that generates an optimal response according to the emotional state of the user based on the analysis results.

[1313] The "means for displaying the generated response" is a function for sending the generated response from the server to the terminal and displaying it so that the person at the counter or the operator can check it.

[1314] This invention relates to a complaint handling support system for customer service representatives and operators. This system provides appropriate responses to complaints, prevents complaints from escalating, and shortens response times. Specific embodiments of this system are described below.

[1315] Hardware and software used

[1316] server

[1317] The server plays a central role in the complaint handling support system. The server uses the following software and hardware:

[1318] Hardware: General cloud server (e.g. AWS EC2 instance)

[1319] software:

[1320] Auto-response generation module (e.g., generative AI model)

[1321] Emotion engines (e.g., natural language processing engines)

[1322] Database (e.g. SQL database)

[1323] Terminal

[1324] A terminal is a device used by a teller or operator. A terminal uses the following hardware and software:

[1325] Hardware: General personal computer (e.g. Windows PC, iPad)

[1326] software:

[1327] Claims Input Interface

[1328] Response Display Interface

[1329] Program processing

[1330] Entering and submitting a claim

[1331] When a user enters a complaint, the terminal sends this information to the server. For example, if a user enters "the product has not arrived," the terminal packages this text data in JSON format and sends it to the server using the HTTPS protocol.

[1332] Claims Analysis

[1333] The server analyzes the text data of the received complaint. Specifically, it uses a generative AI model to perform natural language processing (NLP) to understand the content of the complaint. For example, if a complaint states "the product has not arrived," the server classifies it as a "delivery issue."

[1334] Emotional state analysis

[1335] The server uses an emotion engine to analyze the user's emotional state. It extracts emotions such as anger, frustration, and confusion from the complaint text. For example, for a complaint that "the product has not arrived," the server identifies the emotion "frustrated."

[1336] Generating a response

[1337] The server generates an appropriate response based on the analysis results. Using a generative AI model, it generates a response that reflects the user's emotional state. For example, it generates a response like, "Sorry, we'll check your order. Please wait a moment."

[1338] Sending and Displaying Responses

[1339] The server sends the generated response to the terminal. The terminal displays the response received from the server on its screen. The customer service representative or operator checks this response and takes appropriate action for the user.

[1340] Specific examples

[1341] For example, if a user inputs a complaint stating, "The product has not arrived," the device sends this information to the server. The server then uses a generative AI model to analyze the content of the complaint and an emotion engine to identify the user's emotional state. Based on the analysis results, the server generates a response such as, "We're sorry. We will check your order. Please wait a moment." and sends it to the device. The customer service representative provides this response to the user and attempts to resolve the issue.

[1342] Prompt Sentence Examples

[1343] "If a user enters a complaint such as 'item not received,' how would the server generate a response?"

[1344] In this way, the present invention realizes improved efficiency and quality in handling complaints.

[1345] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1346] Step 1:

[1347] The user enters a claim

[1348] The user inputs a complaint through the interface of the complaint handling support system. For example, the user inputs "The product has not arrived." The input complaint is sent to the terminal as text data. The input data is text that includes the content of the complaint.

[1349] Step 2:

[1350] The device sends the complaint to the server

[1351] The terminal transmits the claims entered by the user to the server in real time. Specifically, the text data of the claims is packaged in JSON format and transmitted to the server using the HTTPS protocol. The input data is the text data of the claims, and the output data is the JSON-formatted data transmitted to the server.

[1352] Step 3:

[1353] The server parses the content of the claim.

[1354] The server analyzes the text data of the received complaint. Specifically, it uses a generative AI model to perform natural language processing (NLP) to understand the content of the complaint. For example, for a complaint that "the product has not arrived," the server classifies it as a "problem with delivery." The input data is the text data of the complaint, and the output data is the classification information of the complaint as a result of the analysis.

[1355] Step 4:

[1356] The server analyzes the user's emotional state

[1357] The server uses an emotion engine to analyze the user's emotional state. Emotions such as anger, frustration, and confusion are extracted from the complaint text. For example, for a complaint that "the product has not arrived," the server identifies the emotion "frustration." The input data is the complaint text data, and the output data is the emotion analysis results.

[1358] Step 5:

[1359] The server generates an appropriate response

[1360] The server generates an appropriate response based on the analysis results. Using a generative AI model, it generates a response that corresponds to the user's emotional state. For example, it generates a response such as, "We're sorry. We will check your order. Please wait a moment." The input data is the complaint classification information and the emotion analysis results, and the output data is the generated response text.

[1361] Step 6:

[1362] The server sends the response to the device

[1363] The server sends the generated response to the terminal. Specifically, it packages the response text data in JSON format and sends it to the terminal using the HTTPS protocol. The input data is the generated response text, and the output data is the JSON-formatted data sent to the terminal.

[1364] Step 7:

[1365] The terminal displays the response

[1366] The terminal displays the response received from the server on the screen. The customer service representative or operator checks this response and takes appropriate action for the user. The input data is the response text sent from the server, and the output data is the response displayed on the screen.

[1367] Step 8:

[1368] The user receives a response

[1369] The user receives a response provided by the agent or operator, which informs the user of the next steps to take toward resolving the problem. The input data is the response provided by the agent or operator, and the output data is the response received by the user.

[1370] (Application example 1)

[1371] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1372] Conventional complaint handling systems had the problem of being unable to properly grasp the emotional state of customers and respond quickly and appropriately. Furthermore, they were unable to predict complaint escalation and take measures in advance, which led to a risk of delayed responses and a decline in customer satisfaction. Furthermore, there was also the problem of insufficient management of response history, meaning that past response history could not be utilized to help with future responses.

[1373] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means. In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further escalation, means for shortening response time, means for analyzing the user's emotional state using an emotion engine, means for generating an appropriate response based on the analysis results, means for predicting the possibility of escalation, and means for managing complaint response history. This makes it possible to properly grasp the customer's emotional state and respond quickly and appropriately. Furthermore, by predicting complaint escalation and taking measures in advance, response delays can be prevented and customer satisfaction can be expected to improve. Furthermore, response history management is enhanced, and past response history can be utilized to help with future responses.

[1374] A "customer service representative" is an employee whose role is to directly receive and respond to inquiries and complaints from customers.

[1375] An "operator" is a staff member whose role is to operate a system or machine and carry out a specific task.

[1376] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[1377] "Means for providing an appropriate response" is a function for generating an appropriate response to a complaint and providing it to the customer.

[1378] "Measures to prevent flame wars" are functions to prevent complaints from spreading and problems from becoming more serious.

[1379] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[1380] An "emotion engine" is a technology for analyzing a user's emotional state.

[1381] The "means for analyzing the user's emotional state" is a function for analyzing the user's emotions using an emotion engine.

[1382] The "means for generating an appropriate response" is a function for generating an appropriate response based on the emotion analysis results.

[1383] "Means for predicting the possibility of escalation" is a function for predicting the possibility of a problem becoming serious based on the content of the complaint and the emotional state.

[1384] The "means for managing complaint handling history" is a function for recording and managing past complaint handling history.

[1385] As an embodiment of the present invention, a complaint handling support system for an online shopping site will be described as an example. This system operates in cooperation with three entities: a server, a terminal, and a user.

[1386] System configuration

[1387] 1. Server:

[1388] Emotion Engine: A software module for analyzing the user's emotional state, specifically using a Python sentiment analysis library (e.g., TextBlob or VADER).

[1389] Generative AI model: An AI model for generating appropriate responses to complaints. Specifically, we use generative AI models such as GPT-3.

[1390] Escalation prediction module: A machine learning model for predicting the likelihood of escalation based on the content of the complaint and the emotional state. Specifically, it uses machine learning libraries such as scikit-learn.

[1391] History management system: A database system for managing complaint handling history. Specifically, it uses a database management system such as MySQL or PostgreSQL.

[1392] 2. Terminal:

[1393] Smartphone app: An application that allows users to input complaints and communicate with the server. This application provides a user interface, allows users to input complaints, displays sentiment analysis results, and receives responses.

[1394] 3. User:

[1395] Customer: Responsible for entering complaints and receiving responses from the system.

[1396] System Operation

[1397] 1. Complaint Reception:

[1398] A user inputs a complaint through a smartphone app, for example, "I'm very unhappy because I didn't receive my product!"

[1399] 2. Emotion analysis:

[1400] The server's emotion engine analyzes the complaint text and determines the user's emotional state. For example, it detects "anger" from the input complaint.

[1401] 3. Auto-response generation:

[1402] The server's generative AI model generates an appropriate response based on the sentiment analysis results, such as "We apologize for the inconvenience. We will investigate this immediately and provide you with a solution."

[1403] 4. Escalation Prediction:

[1404] The server's escalation prediction module predicts the risk of escalation based on the content of the complaint and the emotional state of the customer. For example, it determines the risk as "high risk."

[1405] 5. History Management:

[1406] The server's history management system stores the complaints and their response history in a database, which can be used to handle future complaints.

[1407] Examples and prompts

[1408] Examples:

[1409] Complaint: "I'm very unhappy because I haven't received my item!"

[1410] Sentiment analysis result: Anger

[1411] Automated response: "We apologize for the inconvenience. We will investigate this and provide you with a solution shortly."

[1412] Escalation Risk: High

[1413] Example prompt sentence:

[1414] Customer complaint: "I'm very unhappy because I didn't receive my item!"

[1415] Sentiment analysis result: Anger

[1416] Generate an appropriate response.

[1417] In this way, the server, terminal, and user work together to accurately grasp the emotional state of the customer and respond quickly and appropriately. It also makes it possible to predict complaint escalation and take measures in advance, which prevents delays in response and is expected to improve customer satisfaction. Furthermore, it also improves response history management, making it possible to utilize past response history for future responses.

[1418] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1419] Step 1:

[1420] A user inputs a complaint through a smartphone app. The input complaint text is sent to the server. For example, a complaint such as "I'm very unhappy because my product hasn't arrived!" is input.

[1421] Step 2:

[1422] The emotion engine on the server receives the complaint text and performs emotion analysis. It analyzes the input complaint text and determines the user's emotional state. For example, it uses an emotion analysis library (e.g., TextBlob or VADER) to detect the emotion "anger." The output is the emotional state (e.g., anger).

[1423] Step 3:

[1424] The generative AI model on the server receives the emotion analysis results and generates an appropriate response. Using the emotional state (e.g., anger) and the complaint text as input, a prompt sentence is input to the generative AI model (e.g., GPT-3) to generate a response. For example, a response might be generated that says, "We apologize for the inconvenience. We will investigate this immediately and provide you with a solution." The output is the generated response.

[1425] Step 4:

[1426] The server's escalation prediction module receives the complaint text and emotional state and predicts the risk of escalation. Using the complaint text and emotional state as input, it predicts the escalation risk using a machine learning model (e.g., scikit-learn). For example, it is determined to be "high risk." The output is the escalation risk.

[1427] Step 5:

[1428] The server's history management system stores the complaint and its response history in a database. The complaint text, emotional state, generated responses, and escalation risk are input and stored in a database management system (e.g., MySQL or PostgreSQL). The output is the stored history data.

[1429] Step 6:

[1430] The server sends the generated response to the device. The generated response is sent as input to the user's smartphone app. For example, a response such as "We apologize for the inconvenience. We will investigate this immediately and provide you with a solution" is displayed to the user. The output is the response displayed to the user.

[1431] Step 7:

[1432] The user checks the response through the smartphone app and enters additional complaints or feedback as necessary. If the user checks the response and is satisfied, the complaint handling is completed. If there are additional complaints or feedback, the process starts again from step 1. The output is the user's satisfaction level and additional complaints.

[1433] Example 2

[1434] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1435] In conventional complaint handling systems, it is difficult for the customer service representatives and operators to provide appropriate responses to customer inquiries, placing a heavy mental burden on them. Another issue is that there is a variation in response skills, making it difficult to maintain consistency in service quality. Furthermore, when handling complaints, it is necessary to properly recognize customer emotions and prevent the situation from escalating into a heated argument, but there is a lack of effective means to achieve this.

[1436] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1437] In this invention, the server includes means for generating an appropriate response to an inquiry, means for calculating a confidence score for the generated response, means for displaying the confidence score, means for analyzing the emotional state of the user, means for generating an appropriate response based on the analysis result, means for providing a sense of security, and means for achieving leveling of response skills, thereby enabling tellers and operators to provide consistent, high-quality responses to customers while gaining a sense of security.

[1438] The "means for generating an appropriate response to an inquiry" is a technology for automatically generating an appropriate answer based on the content of an inquiry from a user.

[1439] The "means for calculating a confidence score for a generated response" is a technique for evaluating and quantifying the accuracy and reliability of a generated response.

[1440] A "means for displaying a confidence score" is a technique for visually displaying the calculated confidence score to a user.

[1441] "Means for analyzing the user's emotional state" refers to technology for reading and analyzing emotions from the user's statements and actions.

[1442] "Means for generating an appropriate response based on the analysis results" refers to a technology for generating an optimal response to a user based on the results of emotion analysis.

[1443] "Means for providing psychological security" refers to techniques that enable users to gain a sense of security and trust when using a system.

[1444] "Means for achieving standardization of response skills" refers to technology for reducing the variation in response skills among different users and achieving uniform response quality.

[1445] This invention is a system that enables counter staff and operators to respond appropriately to customer inquiries and complaints. This system uses a generative AI model to generate responses to inquiries and calculates and displays their reliability scores, providing users with a sense of mental security and achieving a leveling of response skills. It also uses an emotion engine to analyze the user's emotional state and generate appropriate responses, improving the quality of complaint handling.

[1446] Hardware and software used

[1447] Server: Provides the computational resources to run generative AI models (e.g., OpenAI's GPT-4).

[1448] Device: The device operated by the user (e.g., PC, tablet, smartphone) displays the response and confidence score sent from the server.

[1449] Emotion engine: Software for analyzing a user's emotional state (e.g., Microsoft's Azure Cognitive Services).

[1450] Program processing

[1451] 1. The user receives a customer inquiry. For example, the customer asks, "How long is the warranty on this product?"

[1452] 2. The server uses the generative AI model to generate an appropriate response based on the prompt, "How long is the warranty on this product?"

[1453] 3. The server calculates a confidence score for the generated response. For example, the generated response might be "This product has a one-year warranty" with a confidence score of 90%.

[1454] 4. The server sends the generated response and confidence score to the device.

[1455] 5. The device displays the received response and the confidence score to the user. For example, it might display "Response: This product has a one-year warranty (confidence score: 90%)" on the screen.

[1456] 6. The user checks the displayed response and confidence score and provides the response to the customer with confidence.

[1457] Specific examples

[1458] The user (operator) receives an inquiry from a customer.

[1459] The server uses a generative AI model to generate a response to the query, "How long is the warranty on this product?"

[1460] The server calculates a confidence score for the generated response "This product has a one-year warranty" and sends the confidence score of 90% to the terminal.

[1461] The terminal displays the response and the confidence score to the user.

[1462] The user can view the confidence score and provide a response to the customer with confidence.

[1463] Prompt Sentence Examples

[1464] "Generate an appropriate response to a customer inquiry such as 'How long is the warranty on this product?'"

[1465] In this way, this system utilizes AI technology to provide psychological security and standardize response skills.

[1466] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1467] Step 1:

[1468] A user receives an inquiry from a customer.

[1469] Input: Customer inquiry (e.g., "How long is the warranty on this product?")

[1470] Specific operation: A user (operator) receives an inquiry from a customer by means of telephone, chat, or the like.

[1471] Step 2:

[1472] The server generates a response using a generative AI model.

[1473] Input: The inquiry received from the user (e.g., "How long is the warranty period for this product?")

[1474] Data processing: Input the query content as a prompt into a generative AI model (e.g., OpenAI's GPT-4).

[1475] Output: The generated response (e.g., "This product has a one-year warranty")

[1476] Specific operation: The server inputs the query content into the generative AI model and generates an appropriate response.

[1477] Step 3:

[1478] The server calculates a confidence score for the generated response.

[1479] Input: The generated response (e.g., "This product has a one-year warranty")

[1480] Data Calculation: An internal algorithm is used to calculate the confidence score of the response.

[1481] Output: Confidence score (e.g. 90%)

[1482] Specific behavior: The server evaluates the accuracy and reliability of the generated response and calculates a confidence score.

[1483] Step 4:

[1484] The server sends the response and the confidence score to the device.

[1485] Input: A generated response (e.g., "This product has a one-year warranty") and a confidence score (e.g., 90%)

[1486] Data processing: Responses and confidence scores are combined into a single data packet.

[1487] Output: Data packets sent to the device

[1488] Specific operation: The server assembles the generated response and confidence score into a data packet and sends it to the terminal.

[1489] Step 5:

[1490] The device displays the response and confidence score to the user.

[1491] Input: Data packet sent by the server (response and confidence score)

[1492] Data processing: Parse the data packets and extract the response and confidence score.

[1493] Output: The response and confidence score displayed to the user (e.g., "Response: This product has a one-year warranty (confidence score: 90%)")

[1494] Specific operation: The terminal analyzes the received data packet and displays the response and confidence score on the screen.

[1495] Step 6:

[1496] The user sees the displayed response and confidence score.

[1497] Input: The response displayed on the device and a confidence score (e.g., "Response: This product has a one-year warranty (confidence score: 90%)")

[1498] Specific operation: The user (operator) checks the displayed response and the reliability score and determines the reliability of the response.

[1499] Step 7:

[1500] The user provides a response to the customer.

[1501] Input: A verified response (e.g., "This product has a one-year warranty")

[1502] Specific operation: The user (operator) responds to the customer based on the confirmed response. For example, he / she tells the customer, "The warranty period for this product is one year."

[1503] In this way, by performing specific actions at each step, the system provides psychological security and equalizes response skills.

[1504] (Application example 2)

[1505] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1506] In conventional complaint handling systems, the front desk staff or operators were unable to properly grasp the user's emotional state, which resulted in delayed responses or inappropriate responses. Furthermore, there was no way to verify the reliability of the responses generated by AI, which often caused the operators to feel uneasy. Furthermore, there was a problem of inconsistency in service quality due to variations in response skills.

[1507] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1508] In this invention, the server includes a means for providing an appropriate response to a complaint, a means for preventing further escalation, a means for shortening response time, a means for displaying a reliability score for the response generated by AI, a means for analyzing the user's emotional state using an emotion engine, and a means for generating an appropriate response based on the analysis results. This allows the counter staff or operators to properly grasp the user's emotional state and provide a reliable response quickly. Furthermore, by providing a sense of mental security and achieving a leveling of response skills, consistency in service quality can be maintained.

[1509] A "customer service representative" is an employee whose role is to directly receive and respond to inquiries and complaints from customers.

[1510] An "operator" is a staff member whose role is to operate a system or machine and carry out a specific task.

[1511] A "complaint handling support system" is a system that provides support for appropriate responses to customer complaints.

[1512] "Means of providing appropriate responses" refers to the function of generating and providing appropriate responses to customer inquiries and complaints.

[1513] "Measures to prevent flame wars" is a function that takes measures to prevent customer dissatisfaction and anger from escalating.

[1514] "Means to shorten response times" refers to functions that enable quick responses to complaints and inquiries.

[1515] "Means for displaying a confidence score for AI-generated responses" is a function that quantifies and displays the confidence of answers generated by AI.

[1516] An "emotion engine" is software or hardware for analyzing a user's emotional state.

[1517] "Means for analyzing the user's emotional state" is a function that reads and analyzes emotions from the user's statements and actions.

[1518] "Means for generating an appropriate response based on the analysis results" is a function that generates the optimal answer based on the results of sentiment analysis.

[1519] "Means for providing mental security" is a function that provides support to enable operators to perform their work with peace of mind.

[1520] "Means to achieve standardization of response skills" refers to a function that reduces the variation in response skills between operators and maintains uniform service quality.

[1521] To implement this invention, the following hardware and software are used: A smartphone is required as hardware, and Python, OpenAI API, and the emotion_recognition library are used as software.

[1522] The server first receives a complaint or inquiry from the user as input. Then, it uses the emotion_recognition library to analyze the user's emotional state. Based on the analysis results, it uses the OpenAI API to generate an AI response to the user input. It then obtains a confidence score for the generated response and displays it to the operator.

[1523] For example, if a user inputs a complaint such as "My security system has been malfunctioning recently and I'm having trouble with it. What should I do?", the emotion analysis result will be "anger." Based on this analysis result, the AI ​​will generate a response such as "We apologize for the inconvenience. Could you please try restarting the system?" and display a confidence score of 0.85.

[1524] An example of a prompt sentence is as follows:

[1525] Please enter your complaint or inquiry: My security system has been malfunctioning lately. What should I do?

[1526] In this way, the server can properly grasp the user's emotional state and provide a reliable response quickly. Furthermore, it provides a sense of security and standardizes response skills, thereby maintaining consistency in service quality.

[1527] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1528] Step 1:

[1529] A user inputs a complaint or inquiry using a smartphone. The input text is sent to the server. The input data is the content of the user's complaint or inquiry.

[1530] Step 2:

[1531] The server passes the received user input to the emotion_recognition library to analyze the user's emotional state. The input data is the text of the user's complaint or inquiry, and the output data is the analyzed emotional state (e.g., "anger," "sadness," "joy," etc.).

[1532] Step 3:

[1533] The server generates an AI response to user input using the OpenAI API based on the analyzed emotional state. The input data is the text of the user's complaint or inquiry and the analyzed emotional state, and the output data is the generated AI response text.

[1534] Step 4:

[1535] The server obtains a confidence score for the generated AI response. The input data is the text of the generated AI response, and the output data is the confidence score (e.g., 0.85).

[1536] Step 5:

[1537] The server displays the generated AI response and confidence score on the operator's smartphone. The input data is the text of the generated AI response and the confidence score, and the output data is the information displayed on the operator's smartphone.

[1538] Step 6:

[1539] The operator reviews the displayed AI response and confidence score and takes appropriate action for the user if necessary. The input data is the text of the AI ​​response that the operator reviews and the confidence score, and the output data is the final response that the operator provides to the user.

[1540] In this way, the server can properly grasp the user's emotional state and provide a reliable response quickly. Furthermore, it provides a sense of security and standardizes response skills, thereby maintaining consistency in service quality.

[1541] Example 3

[1542] Next, a description will be given of Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1543] With conventional complaint handling systems, it was difficult for the front desk staff and operators to respond appropriately to complaints, resulting in inconsistent response quality. Furthermore, it was not possible to properly grasp the user's emotions when handling complaints, making it difficult to prevent complaints from escalating into flames. Furthermore, there was insufficient effort to improve the skills of operators or maintain their motivation, leading to issues with continued employment. To solve these problems, a system was needed that would improve the quality of complaint handling, shorten response times, and help operators improve their skills and maintain their motivation.

[1544] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[1545] In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further flaming, means for realizing a reduction in response time, means for analyzing a user's emotional state using an emotion analysis engine, means for generating an appropriate response using a generative AI model, means for evaluating the quality of the generated response, means for providing feedback to an operator based on the evaluation results, and means for supporting the improvement of the operator's skills. This makes it possible to improve the quality of complaint handling, shorten response time, and support the improvement of the operator's skills and maintenance of motivation.

[1546] A "customer service representative" is an employee whose role is to directly receive and respond to inquiries and complaints from customers.

[1547] An "operator" is a person who operates the system and handles customer service and complaints.

[1548] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[1549] "Means for providing appropriate responses" is a function for generating and providing appropriate responses to customer complaints.

[1550] "Flame prevention measures" are functions that allow for the proper handling of customer dissatisfaction and anger and prevent the problem from escalating.

[1551] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[1552] An "emotion analysis engine" is software or hardware for analyzing a customer's emotional state and providing the results.

[1553] A "generative AI model" is an artificial intelligence model that generates an appropriate response based on a given prompt.

[1554] The "means for assessing the quality of the response" is a function for assessing the appropriateness and effectiveness of the generated response.

[1555] The "means for providing feedback" is a function for providing the operator with advice and points for improvement based on the results of the response quality evaluation.

[1556] "Means to support skill improvement" is a function that provides training and educational programs to improve the response skills of operators.

[1557] "Means for providing a sense of mental security" refers to a function that reduces the stress and anxiety that operators feel while handling complaints, and provides a sense of security.

[1558] "Means to achieve standardization of response skills" refers to a function that reduces the variation in response skills between operators and maintains a consistent level of quality.

[1559] "Means to support the resolution of employment continuity issues" refers to functions to reduce the turnover rate of operators and support long-term employment continuity.

[1560] This invention is a support system for enabling counter staff and operators to appropriately handle customer complaints. The purpose of this system is to improve the quality of complaint handling, shorten response times, and support the improvement of operator skills and maintenance of motivation.

[1561] Hardware and software used

[1562] Hardware: Servers, devices (PCs, tablets, smartphones)

[1563] Software: Generative AI models (e.g., GPT-4), sentiment analysis engines (e.g., Affectiva), database management systems (e.g., MySQL)

[1564] Specific operation of the system

[1565] Receiving a complaint

[1566] User: A customer service representative or operator receives a complaint from a customer.

[1567] Terminal: The counter staff or operator enters the details of the complaint into the terminal. For example, if a customer complains that the product has not arrived, the details are entered in text format.

[1568] Emotion analysis

[1569] Terminal: Sends the entered complaint details to the server.

[1570] Server: Receives complaint content and sends it to the sentiment analysis engine.

[1571] Sentiment analysis engine: Analyzes the content of complaints and identifies the user's emotional state. For example, it can be analyzed as "anger."

[1572] Server: Receives the analysis results from the sentiment analysis engine and proceeds to the next processing step.

[1573] Response Generation

[1574] Server: Based on the results of the sentiment analysis, the server sends a prompt to the generative AI model. For example, it sends a prompt such as, "The customer is angry because the product has not arrived. Please generate an appropriate response."

[1575] Generative AI model: Generates an appropriate response based on the prompt, for example, "Dear customer, we are very sorry. We will investigate this immediately and provide a solution as soon as possible."

[1576] Server: Receives the generated response and proceeds to the next processing step.

[1577] Response quality assessment

[1578] Server: Runs algorithms to evaluate the quality of the generated responses, e.g., their appropriateness, impact on customer satisfaction, etc.

[1579] Server: Stores the evaluation results in a database and generates feedback.

[1580] Feedback and skill development support

[1581] Server: Generates feedback to the operator based on the evaluation results. For example, "Your response was appropriate, but it would be better if you provided a more specific solution."

[1582] Terminal: Operators receive feedback and undergo training programs to improve their skills, for example, through simulation training to improve the quality of their responses.

[1583] Examples of concrete examples and prompts

[1584] Example: A customer receives a complaint that the product has not arrived.

[1585] Terminal: The counter staff enters the details of the complaint.

[1586] Server: Sends the complaint content to the sentiment analysis engine.

[1587] Sentiment analysis engine: Analyzes the customer's emotional state as "anger."

[1588] Server: Sends the prompt statement "The customer is angry because the product has not arrived" to the generative AI model.

[1589] Generative AI model: Generates the response, "Dear customer, we are very sorry. We will investigate this immediately and provide a solution as soon as possible."

[1590] Server: Evaluates the quality of the response and provides feedback.

[1591] Example prompt sentence:

[1592] "A customer is upset that their item hasn't arrived. Generate an appropriate response."

[1593] "A customer is unhappy with the quality of service. Please generate an appropriate response."

[1594] This system improves the skills of operators and maintains their motivation, thereby resolving the issue of continued employment. The flow of the identification process in the third embodiment will be described with reference to FIG.

[1595] Step 1: Receiving the claim

[1596] User: A customer service representative or operator receives a complaint from a customer. For example, a customer complains that the product has not arrived.

[1597] Terminal: The counter staff or operator enters the details of the claim into the terminal. The entered details of the claim are saved in text format.

[1598] Input: Customer complaint (e.g. "Product not received").

[1599] Output: Claims data in text format.

[1600] Step 2: Sentiment Analysis

[1601] Terminal: Sends the entered complaint details to the server.

[1602] Server: Receives complaint content and sends it to the sentiment analysis engine.

[1603] Sentiment analysis engine: Analyzes the content of complaints and identifies the user's emotional state. For example, it can be analyzed as "anger."

[1604] Input: Claim data in text format.

[1605] Output: Sentiment analysis result (e.g. "Anger").

[1606] Step 3: Response Generation

[1607] Server: Based on the results of the sentiment analysis, the server sends a prompt to the generative AI model. For example, it sends a prompt such as, "The customer is angry because the product has not arrived. Please generate an appropriate response."

[1608] Generative AI model: Generates an appropriate response based on the prompt, for example, "Dear customer, we are very sorry. We will investigate this immediately and provide a solution as soon as possible."

[1609] Input: Sentiment analysis results and prompt sentence.

[1610] Output: The generated response sentence.

[1611] Step 4: Response quality assessment

[1612] Server: Runs algorithms to evaluate the quality of the generated responses, e.g., their appropriateness, impact on customer satisfaction, etc.

[1613] Server: Stores the evaluation results in a database and generates feedback.

[1614] Input: The generated response sentence.

[1615] Output: Response quality evaluation results and feedback.

[1616] Step 5: Feedback and skill development support

[1617] Server: Generates feedback to the operator based on the evaluation results. For example, "Your response was appropriate, but it would be better if you provided a more specific solution."

[1618] Terminal: Operators receive feedback and undergo training programs to improve their skills, for example, through simulation training to improve the quality of their responses.

[1619] Input: Response quality assessment results and feedback.

[1620] Output: Upskilling of operators and implementation of training programs.

[1621] (Application example 3)

[1622] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1623] Conventional complaint handling systems provided insufficient support for counter staff and operators to respond appropriately to complaints, resulting in inconsistent response quality. Furthermore, they lacked the ability to properly analyze user emotions and generate responses based on those emotions, which led to complaints being handled inappropriately, sometimes resulting in outrage. Furthermore, the mental burden on operators was heavy, raising concerns about continued employment. To solve these issues, a system that analyzes user emotions and generates appropriate responses was needed.

[1624] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[1625] In this invention, the server includes means for providing an appropriate response to a complaint, means for preventing further flaming, means for shortening response time, means for analyzing a user's emotions, means for generating an appropriate response based on the analysis results, means for evaluating the quality of the generated response, means for feeding back the evaluation results, means for providing a sense of mental security, means for leveling response skills, and means for supporting the resolution of issues related to continued employment. This enables the front desk staff or operators to respond appropriately according to the user's emotions, improves the quality of complaint handling, prevents flaming, reduces mental burden, and promotes continued employment.

[1626] A "contact person" is a staff member whose role is to directly receive and respond to inquiries and complaints from customers and users.

[1627] An "operator" is a staff member whose role is to operate a system or machine and carry out a specific task.

[1628] A "complaint handling support system" is a system that supports appropriate responses to complaints from customers and users.

[1629] "Means for providing appropriate responses" refers to the function for generating and providing appropriate responses to complaints.

[1630] "Measures to prevent flame wars" is a function to prevent complaints from being handled inappropriately and the problem from escalating.

[1631] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[1632] "Means for analyzing user emotions" is a function for analyzing emotions from user comments and actions.

[1633] The "means for generating an appropriate response" is a function for generating an appropriate response based on the analyzed emotion.

[1634] The "means for evaluating the quality of the generated response" is a function for evaluating the quality of the generated response.

[1635] The "means for feeding back evaluation results" is a function for feeding back the quality of the evaluated response to the operator.

[1636] The "means for providing a sense of mental security" is a function for providing a sense of mental security to the operator.

[1637] The "means for achieving equalization of response skills" is a function for equalizing the response skills of the agents.

[1638] "Means to support the resolution of issues related to continued employment" refers to functions that support the improvement of operators' skills and maintenance of motivation, thereby resolving issues related to continued employment.

[1639] To implement this invention, it is necessary to build a complaint handling support system. This system supports counter staff and operators in appropriately handling complaints, and uses the following hardware and software.

[1640] Hardware

[1641] Server: Serves as the core of the complaint handling support system. It processes and analyzes data and generates responses.

[1642] Terminal: A smartphone or computer used by a teller or operator that receives user input and displays responses from the server.

[1643] software

[1644] OpenAI API: Provides the generative AI model used to generate responses.

[1645] EmotionEngine: A library for analyzing user emotions.

[1646] Database: A system for storing claims and response data.

[1647] Processing flow

[1648] 1. Receiving user input

[1649] The user inputs a complaint through the terminal, for example, "The security alarm keeps ringing and I'm having trouble."

[1650] 2. Emotion analysis

[1651] The server uses the Emotion Engine to analyze emotions from user input. For example, it recognizes the emotion "confused" from the user input.

[1652] 3. Response Generation

[1653] The server generates an appropriate response using the OpenAI API based on the analyzed emotion. An example prompt is as follows:

[1654] User Emotion: Confused

[1655] User Input: My security alarm keeps going off and I'm having trouble.

[1656] Generate an appropriate response.

[1657] 4. Response Quality Assessment

[1658] The server evaluates the quality of the generated response, for example by checking whether it contains keywords such as "thank you" or "appreciation."

[1659] 5. Providing Feedback

[1660] The server then provides feedback on the evaluation results to the operator, allowing the operator to improve their own response skills.

[1661] Specific examples

[1662] If a user types "I'm bothered by the security alarm that keeps going off," the server uses the Emotion Engine to analyze the emotion "confused," then uses the OpenAI API to generate a response like this:

[1663] "I see you're having trouble. We'll take care of the security alarm issue right away. Please wait a moment."

[1664] By evaluating the quality of the generated response and feeding back the evaluation results to the operator, the operator's response skills can be improved.

[1665] In this way, the customer service representative or operator can respond appropriately based on the user's emotions, improving the quality of complaint handling, preventing outrage, reducing mental stress, and promoting continued employment.

[1666] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[1667] Step 1:

[1668] The user inputs a complaint through the terminal. For example, the input complaint might be something like, "My security alarm keeps ringing and I'm having trouble." This input data is sent to the server.

[1669] Step 2:

[1670] The server passes the received user input data to the Emotion Engine, which analyzes the user's emotions. The Emotion Engine analyzes the input data and recognizes the emotion, for example, "confusion." The analysis results are returned to the server.

[1671] Step 3:

[1672] The server generates an appropriate response using the OpenAI API based on the analyzed emotion. The server generates a prompt sentence like the following and sends it to the OpenAI API:

[1673] User Emotion: Confused

[1674] User Input: My security alarm keeps going off and I'm having trouble.

[1675] Generate an appropriate response.

[1676] The OpenAI API generates a response based on this prompt and returns it to the server.

[1677] Step 4:

[1678] The server evaluates the quality of the generated response, for example by checking whether the response contains keywords such as "thank you" or "appreciation," and stores the evaluation results locally.

[1679] Step 5:

[1680] The server then feeds back the evaluation results to the operator, who can then display the feedback on their terminal, allowing them to improve their own response skills.

[1681] Step 6:

[1682] The operator responds appropriately to the user based on the feedback provided by the server, for example, "I see you're having trouble. We'll get back to you about your security alarm issue. Please wait a moment."

[1683] In this way, the customer service representative or operator can respond appropriately based on the user's emotions, improving the quality of complaint handling, preventing outrage, reducing mental stress, and promoting continued employment.

[1684] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1685] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1686] Another example of generative AI is Gemini (internet search engine). <url: https: gemini.google.com ?hl="ja">) are listed.

[1687] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[1688] [Third embodiment]

[1689] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1690] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[1691] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1692] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[1693] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1694] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1695] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1696] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1697] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1698] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1699] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1700] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[1701] "Example 1"

[1702] A first embodiment of the present invention is a complaint handling support system for front desk staff and operators. This system is equipped with an automatic response generation module using AI (artificial intelligence) as a means of providing appropriate responses to complaints. This module analyzes the content of the complaint and generates an appropriate response. It also has a function to predict the escalation of the complaint as a means of preventing further escalation. Furthermore, as a means of shortening response time, an optimization algorithm is used to speed up response generation. "Example 2"

[1703] A second embodiment of the present invention is a system further including a means for providing a sense of mental security and a means for achieving the standardization of response skills. As a means for providing a sense of mental security, the system is provided with a function for displaying a reliability score for the response generated by the AI. This allows the operator to confirm the reliability of the response proposed by the AI ​​and achieve a sense of mental security. Furthermore, as a means for achieving the standardization of response skills, the operator can refer to the response generated by the AI, thereby reducing the variation in response skills between operators.

[1704] "Example 3"

[1705] A third embodiment of the present invention is a system that further includes a means for supporting the resolution of the issue of continued employment. Specifically, the system is equipped with a function for evaluating the quality of responses generated by AI and supporting the improvement of the operator's skills. This helps improve the operator's skills and maintain their motivation, thereby supporting the resolution of the issue of continued employment.

[1706] The processing flow of each embodiment will be described below.

[1707] "Example 1"

[1708] Step 1: The counter clerk or operator receives the complaint.

[1709] Step 2: The system analyzes the content of the claim.

[1710] Step 3: An automatic response generation module using AI (artificial intelligence) generates an appropriate response.

[1711] Step 4: The system predicts claim escalation.

[1712] Step 5: An optimization algorithm is implemented to speed up response generation, reducing response time.

[1713] "Example 2"

[1714] Step 1: The teller or operator reviews the response generated by the AI.

[1715] Step 2: The system displays a confidence score for the AI-generated response.

[1716] Step 3: The teller or operator decides on the final response based on the reliability score.

[1717] Step 4: Counter staff and operators use the responses generated by the AI ​​as a reference to standardize their response skills.

[1718] "Example 3"

[1719] Step 1: The teller / operator uses the AI-generated response.

[1720] Step 2: The system evaluates the quality of the AI-generated response.

[1721] Step 3: The system provides feedback to help operators improve their skills.

[1722] Step 4: Counter staff and operators use the feedback to improve their skills and maintain motivation.

[1723] Example 1

[1724] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1725] Conventional complaint handling systems did not provide sufficient support for the front desk staff or operators to respond appropriately to complaints, which often resulted in variations in the quality and speed of responses. In addition, there was no way to predict the risk of a complaint escalating, and a delayed response was likely to escalate the situation. Furthermore, with the need to reduce response times, it was difficult to generate a prompt and appropriate response. A system that could solve these issues was needed.

[1726] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1727] In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further escalation, means for shortening response time, means for analyzing the content of the complaint, means for generating automatic responses, means for predicting escalation, and means for accelerating response generation. This improves the quality and speed of complaint handling, reduces the risk of escalation, and makes it possible to provide prompt and appropriate responses.

[1728] The "means for providing an appropriate response to a complaint" is a function for generating and providing an appropriate response to a complaint from a user.

[1729] "Measures to prevent further escalation" is a function to prevent complaints from escalating and the problem from expanding.

[1730] The "means for realizing a reduction in response time" is a function for reducing response time by quickly generating and providing responses to complaints.

[1731] The "means for analyzing the content of a complaint" is a function for analyzing the content of a complaint from a user and identifying the type of problem and the emotion behind it.

[1732] The "means for generating an automatic response" is a function for automatically generating an appropriate response based on the analysis results.

[1733] "Means for predicting escalation" is a function for predicting the possibility of a claim escalating and assessing the risk.

[1734] The "means for speeding up response generation" is a function for optimizing the response generation process and providing a response quickly.

[1735] This invention is a support system for enabling counter staff and operators to respond appropriately to complaints. This system includes multiple means for providing appropriate responses to complaints, preventing flaming, and shortening response times.

[1736] System configuration

[1737] Hardware and Software

[1738] Server: Analyzes complaint content, generates automatic responses, predicts escalations, and speeds up response generation.

[1739] Terminal: The device (PC, smartphone, etc.) through which the user enters claims and receives responses from the server.

[1740] Natural language processing (NLP) technology: Used to analyze the content of claims, specifically using Google Cloud Natural Language API and IBM Watson Natural Language Understanding.

[1741] Generative AI model: Used to generate automated responses. Specifically, we use OpenAI's GPT-3.

[1742] Machine learning algorithms: Used to perform escalation predictions, specifically using scikit-learn and TensorFlow.

[1743] Caching techniques and parallel processing are used to speed up response generation.

[1744] System Operation

[1745] 1. Entering a claim

[1746] The user inputs the details of the complaint from the terminal.

[1747] Example: User types "The product is of poor quality."

[1748] 2. Analysis of the complaint

[1749] The server receives the complaint sent by the user.

[1750] The server parses the claim content using the Google Cloud Natural Language API.

[1751] Specifically, the text of the complaint is sent to an API, which performs sentiment analysis and keyword extraction on the text.

[1752] Example: Extract the keyword "quality issue" from the text "poor product quality" and determine the sentiment as "negative."

[1753] 3. Generate Auto-Responses

[1754] The server generates an appropriate response using a generative AI model (GPT-3) based on the analysis results.

[1755] Specifically, the analysis results are input into a generative AI model as a prompt sentence, and a response sentence is generated.

[1756] Example: Enter the prompt text "The user has entered a complaint that the product quality is poor. Please generate an appropriate response." and get the generated response "I'm sorry. Can you tell me more about the quality of the product?"

[1757] 4. Escalation Prediction

[1758] The server uses machine learning algorithms to predict the likelihood of a claim escalation.

[1759] Specifically, we use scikit-learn to build a model that evaluates the risk of escalation based on past claim data and calculate a risk score for current claims.

[1760] Example: A high risk score is assigned to a complaint that has not been resolved despite repeated contact.

[1761] 5. Providing a Response

[1762] The server provides the generated response to the user.

[1763] Specifically, the generated response sentence is sent to the terminal and displayed on the user's screen.

[1764] Example: "Sorry, can you tell me more about the quality of your product?" is the response displayed on the user's device.

[1765] 6. Faster response generation

[1766] The server uses caching techniques and parallel processing to speed up response generation.

[1767] Specifically, past responses to the same type of claim are cached and reused, and processing speed is improved by processing multiple claims in parallel.

[1768] Example: If the response to a complaint about "poor quality of product" is stored in the cache, retrieve the response from the cache and serve it quickly.

[1769] Specific examples

[1770] The user inputs a complaint saying that the product is of poor quality.

[1771] The server sends the complaint content to the Google Cloud Natural Language API and extracts the keyword "quality issue" and the sentiment "negative."

[1772] The server inputs the prompt text "The user has input a complaint that 'the quality of the product is poor.' Please generate an appropriate response." into the generative AI model (GPT-3), which generates the response "I'm sorry. Could you tell me more about the quality of the product?"

[1773] The server calculates the risk score for escalation using scikit-learn and assigns a high risk score.

[1774] The server sends the generated response to the user's terminal and displays it on the user's screen.

[1775] The server caches responses to the same type of claim to speed up subsequent response generation.

[1776] The above is an embodiment of the present invention.

[1777] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1778] Step 1:

[1779] Entering a claim

[1780] The user inputs the details of the complaint from the terminal.

[1781] Input: The complaint entered by the user (e.g., "The product is of poor quality").

[1782] Output: The claim is sent to the server.

[1783] Specific operation: The user enters the details of the complaint into the input form on the terminal and presses the send button.

[1784] Step 2:

[1785] Receiving complaint details

[1786] The server receives the complaint sent by the user.

[1787] Input: The complaint sent by the user.

[1788] Output: The claim is saved on the server.

[1789] Specific operation: The server receives the HTTP request and saves the claim details in the database.

[1790] Step 3:

[1791] Analysis of complaints

[1792] The server parses the claim content using the Google Cloud Natural Language API.

[1793] Input: Saved claim details.

[1794] Output: Analysis results (keywords, sentiment analysis results, etc.).

[1795] Specific operation: The server sends the claim content to the API and obtains the analysis results returned by the API.

[1796] Step 4:

[1797] Generate an auto-response

[1798] The server generates an appropriate response using a generative AI model (GPT-3) based on the analysis results.

[1799] Input: Analysis results.

[1800] Output: The generated response sentence.

[1801] Specific operation: The server inputs the analysis result as a prompt sentence into the generative AI model and obtains the generated response sentence.

[1802] Example: Enter the prompt text "The user has entered a complaint that the product quality is poor. Please generate an appropriate response." to generate the response "I'm sorry. Can you tell me more about the quality of your product?"

[1803] Step 5:

[1804] Escalation Prediction

[1805] The server uses machine learning algorithms to predict the likelihood of a claim escalation.

[1806] Input: Claim details and analysis results.

[1807] Output: Escalation risk score.

[1808] What happens: The server uses a model built from past claims data to calculate a risk score for the current claim.

[1809] Example: A high risk score is assigned to a complaint that has not been resolved despite repeated contact.

[1810] Step 6:

[1811] Providing a response

[1812] The server provides the generated response to the user.

[1813] Input: The generated response sentence.

[1814] Output: The response is displayed on the user's terminal.

[1815] Specific operation: The server sends the generated response to the user's terminal and displays it on the user's screen.

[1816] Example: "Sorry, can you tell me more about the quality of your product?" is the response displayed on the user's device.

[1817] Step 7:

[1818] Faster response generation

[1819] The server uses caching techniques and parallel processing to speed up response generation.

[1820] Input: Past complaints and responses.

[1821] Output: The cached response.

[1822] What it does: The server caches and reuses past responses to the same type of claim, and processes multiple claims in parallel to improve processing speed.

[1823] Example: If the response to a complaint about "poor quality of product" is stored in the cache, retrieve the response from the cache and serve it quickly.

[1824] (Application example 1)

[1825] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1826] With conventional complaint handling systems, customer service representatives and operators had to manually handle complaints, resulting in long response times and lower customer satisfaction. It was also difficult to predict the escalation of complaints, meaning that it was not possible to notify superiors at the appropriate time, increasing the risk of complaints escalating. Furthermore, there was a large variation in response skills and a significant mental burden, which led to issues with continued employment.

[1827] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1828] In this invention, the server includes a means for providing appropriate responses to complaints, a means for preventing further escalation, a means for shortening response time, a means for analyzing the content of the complaint and generating an appropriate response, a means for predicting the escalation of the complaint, a means for automatically notifying a superior, a means for generating responses using a generative AI model, and a means for generating responses using prompt sentences. This enables efficient and rapid complaint handling, improving customer satisfaction and reducing the risk of complaints escalating. It also solves the issue of continued employment by leveling out response skills and providing psychological security.

[1829] A "contact person" is a person in charge of directly responding to inquiries and complaints from customers.

[1830] An "operator" is a person who operates a system or machine and is responsible for carrying out a specific task.

[1831] A "complaint handling support system" is a system that supports appropriate responses to customer complaints.

[1832] "Means for providing appropriate responses" refers to the function for generating and providing appropriate responses to complaints.

[1833] "Measures to prevent flame wars" are functions to prevent complaints from spreading and problems from becoming more serious.

[1834] "Means for reducing response time" refers to functions for reducing the time it takes to respond to complaints.

[1835] The "means for analyzing the content of the claim and generating an appropriate response" is a function for analyzing the content of the claim and generating an appropriate response based on that analysis.

[1836] The "means for predicting the escalation of a complaint" is a function for predicting the possibility that a complaint will be escalated to a higher-level person in charge.

[1837] The "means for automatically notifying a superior" is a function for automatically notifying a superior when a complaint escalates.

[1838] "Response generation means using a generative AI model" is a function for generating responses to claims using an artificial intelligence model.

[1839] The "means for generating a response using a prompt sentence" is a function for generating a response to a complaint using a specific input sentence (prompt sentence).

[1840] As an example of an embodiment of this invention, we will explain a customer support application for an online shopping site. This system uses AI to automatically generate appropriate responses to customer complaints and inquiries, enabling quick responses. It also has a function to predict complaint escalation and automatically notify higher-level personnel as necessary.

[1841] System Configuration

[1842] The system consists of the following major components:

[1843] 1. Complaint Analysis Module: Analyzes the content of the complaint and generates an appropriate response.

[1844] 2. Escalation prediction module: predicts the escalation of claims.

[1845] 3. Notification module: Automatically notify superiors when escalation is predicted.

[1846] 4. Generative AI model: An artificial intelligence model for generating responses to claims.

[1847] 5. Prompt generation module: Generates a response using a specific input sentence (prompt sentence).

[1848] Hardware and Software

[1849] The system uses an internet-enabled smartphone or PC as hardware, and Python and OpenAI API as software.

[1850] Processing flow

[1851] 1. Entering the details of the complaint: The user enters the details of the complaint.

[1852] 2. Response generation: The claim analysis module analyzes the content of the claim and generates an appropriate response using a generative AI model.

[1853] 3. Escalation Prediction: The escalation prediction module predicts the likelihood of a claim escalation.

[1854] 4. Notification: When an escalation is predicted, the notification module automatically notifies superiors.

[1855] Specific examples

[1856] For example, if a customer enters a complaint stating, "I did not receive the product, so I would like a refund," the system operates as follows:

[1857] 1. Entering the complaint details: The user enters, "I did not receive the product, so I would like a refund."

[1858] 2. Response generation: The complaint analysis module analyzes this content and uses a generative AI model to generate a response such as, "We apologize for the inconvenience. We will immediately investigate the issue of the non-delivery of the product and proceed with the refund process."

[1859] 3. Escalation Prediction: The escalation prediction module predicts that the likelihood of escalation for this claim is "high."

[1860] 4. Notification: The notification module automatically notifies the superior, who then takes over the response.

[1861] Prompt Sentence Examples

[1862] Please enter your complaint: I did not receive my item and would like a refund.

[1863] In this way, the efficiency of customer support on the online shopping site can be improved, and customer satisfaction can be improved.

[1864] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1865] Step 1:

[1866] The user inputs the complaint details.

[1867] Input: The user enters, "I never received my item and would like a refund."

[1868] Specific operation: The user enters the content of the complaint in text format into an input form on a smartphone or PC.

[1869] Step 2:

[1870] The server receives the complaint content and sends it to the complaint analysis module.

[1871] Input: The claim entered by the user.

[1872] Output: The claim content sent to the claim analysis module.

[1873] Specific operation: The server receives input from the user and sends the content to the complaint analysis module.

[1874] Step 3:

[1875] The claim analysis module analyzes the content of the claim and generates an appropriate response using a generative AI model.

[1876] Input: The claim content sent to the claim analysis module.

[1877] Output: The generated response text.

[1878] Specific operation: The claim analysis module analyzes the content of the claim using natural language processing technology and sends it as a prompt to the generative AI model. The generative AI model generates an appropriate response and returns it to the claim analysis module.

[1879] Step 4:

[1880] The server generates a response and sends it to the user.

[1881] Input: The generated response text.

[1882] Output: The response text that is displayed to the user.

[1883] Specific operation: The server receives the generated response and sends it to the user's device, which displays the response.

[1884] Step 5:

[1885] An escalation prediction module predicts the likelihood of a claim escalation.

[1886] Input: Claim details.

[1887] Output: Escalation prediction (high / low).

[1888] Specific operation: The escalation prediction module analyzes the content of the complaint and predicts the possibility of escalation using a generative AI model. The prediction result is returned to the server.

[1889] Step 6:

[1890] The server receives the predicted results of the escalation and automatically notifies the superior as necessary.

[1891] Input: The predicted outcome of the escalation.

[1892] Output: Notification to superiors.

[1893] Specific operation: The server receives the escalation prediction result, and if the prediction result is "high," it sends a notification to the superior via email or a dedicated application.

[1894] In this way, the system can make complaint handling more efficient and faster, improving customer satisfaction and reducing the risk of complaints escalating.

[1895] Example 2

[1896] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1897] Conventional complaint handling support systems do not provide sufficient support for operators to respond appropriately, and lack means to provide psychological security or to standardize response skills. They also lack effective means to shorten response times or prevent complaints from escalating. This increases the burden on operators and leads to inconsistent response quality, which is an issue.

[1898] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1899] In this invention, the server includes means for receiving an inquiry, means for generating a response using a generative AI model, means for calculating a reliability score for the generated response, means for displaying the response and the reliability score, means for providing an appropriate response to the complaint, means for preventing further flaming, and means for realizing a reduction in response time. This allows the operator to confirm the reliability of the response generated by the AI ​​and gain psychological peace of mind, while also reducing the variation in response skills, shortening response time, and preventing flaming.

[1900] The "means for receiving an inquiry" is a function that enables the server to receive an inquiry from a user.

[1901] "Means for generating a response using a generative AI model" is a function for generating an appropriate response to a received inquiry using a generative AI model.

[1902] The "means for calculating a confidence score for a generated response" is a function for evaluating the reliability of a generated response and calculating a confidence score.

[1903] The "means for displaying the response and the confidence score" is a function for visually displaying the generated response and its confidence score to the user.

[1904] "Means for providing appropriate responses to complaints" refers to a function for providing responses that allow appropriate responses to be made to complaints.

[1905] "Measures to prevent further escalation" is a function that provides preventative measures to prevent problems from escalating when dealing with complaints.

[1906] "Means to shorten response times" refers to functions that improve efficiency in order to shorten the time it takes to respond to complaints.

[1907] "Means for providing psychological reassurance" is a function that allows operators to gain psychological reassurance by confirming the reliability of the AI's responses.

[1908] "Means for achieving standardization of response skills" refers to a function for reducing the variation in response skills among operators and achieving uniform response.

[1909] The "means for supporting the resolution of issues regarding employment continuity" is a function for providing support for the resolution of issues regarding the employment continuity of operators.

[1910] This invention is a complaint handling support system that uses a generative AI model to generate responses and displays their reliability scores, thereby providing peace of mind to operators and achieving a leveling of response skills. Specific embodiments of this system are described below.

[1911] Server Processing

[1912] The server receives a query from the user. The query is a prompt sentence, such as "Please tell me the progress of the new project." The server inputs this query into a generative AI model (e.g., GPT-4) and generates an appropriate response. For the generated response, the server calculates a confidence score. This confidence score is calculated using the AI ​​model's internal evaluation mechanism or an additional evaluation algorithm. The server sends the generated response and confidence score to the terminal.

[1913] Terminal handling

[1914] The device receives the response and confidence score sent from the server. The received response and confidence score are displayed on the device screen. For example, it may say, "Current progress is going well. The next milestone is next week. (Confidence score: 85%)." The device makes it easy to understand visually by displaying a high confidence score in green and a low confidence score in red.

[1915] User Action

[1916] The user checks the response and the confidence score displayed on the device. The confidence score is used to evaluate the reliability of the response. For example, if the confidence score is 85%, the user can trust the response and adopt it. If the confidence score is low, the user can make a second inquiry. When making a second inquiry, the user can add more specific questions by referring to the previous response.

[1917] Specific examples

[1918] Example 1: Means of providing psychological security

[1919] The user sends a query such as "How is the progress of the new project going?" The server uses a generative AI model (e.g., GPT-4) to generate a response such as "Current progress is going well. The next milestone is next week," and calculates a confidence score of 85%. The device displays this response and the confidence score, and the user can check the confidence score and feel reassured.

[1920] Example 2: A means to achieve leveling of response skills

[1921] Operator A receives an inquiry saying, "What are the steps to take in handling a complaint?" The server uses a generative AI model to generate a response saying, "First, listen carefully to the customer, then confirm the details of the problem, and finally propose a solution." Operator A responds using this response as a reference, and Operator B can also refer to the same steps when handling similar inquiries, thereby reducing the variability in response skills.

[1922] Prompt Sentence Examples

[1923] "Please tell me how your new project is progressing."

[1924] "Please tell me the procedure for handling complaints."

[1925] These prompts can be fed into a generative AI model to generate an appropriate response and confidence score.

[1926] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1927] Step 1:

[1928] The server receives a query from a user. When the user inputs a prompt such as "Please tell me the progress of the new project" through the terminal and presses the send button, the server receives the query. The input is the user's query, and the output is the query data stored on the server.

[1929] Step 2:

[1930] The server inputs the received query into the generative AI model. Specifically, the server sends the query content to the generative AI model (e.g., GPT-4) via an API. The input is the query data, and the output is the prompt sent to the generative AI model.

[1931] Step 3:

[1932] The generative AI model generates a response based on the prompt sentence. The generative AI model uses an internal natural language processing algorithm to generate an appropriate response. The input is the prompt sentence, and the output is the generated response text.

[1933] Step 4:

[1934] The server calculates a confidence score for the generated response. The server evaluates the reliability of the response using the AI ​​model's internal evaluation mechanism or additional evaluation algorithms, and calculates the confidence score. The input is the generated response text, and the output is the confidence score.

[1935] Step 5:

[1936] The server sends the generated response and confidence score to the terminal. The server combines the response text and confidence score into a single data packet and sends it to the terminal. The input is the response text and confidence score, and the output is the data packet sent to the terminal.

[1937] Step 6:

[1938] The terminal receives the response and confidence score sent from the server. The terminal analyzes the received data packet and extracts the response text and confidence score. The input is the data packet sent from the server, and the output is the analyzed response text and confidence score.

[1939] Step 7:

[1940] The terminal displays the received response and confidence score to the user. The terminal screen displays "Current progress is good. The next milestone is next week. (Confidence score: 85%)." The input is the parsed response text and confidence score, and the output is the information displayed to the user.

[1941] Step 8:

[1942] The user checks the response and the confidence score displayed on the terminal. The user evaluates the reliability of the response based on the confidence score. The input is the displayed response text and the confidence score, and the output is the user's evaluation result.

[1943] Step 9:

[1944] The user can either adopt the response or ask again. If the confidence score is high, the user can trust the response and adopt it. If the confidence score is low, the user can ask again with more specific questions. The input is the user's evaluation result, and the output is the adoption of a new query or response.

[1945] (Application example 2)

[1946] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1947] In conventional complaint response support systems, operators had to provide appropriate responses to complaints, which placed a heavy mental burden on them, and their response skills often varied. Furthermore, when security operators received AI suggestions, they lacked the means to verify their reliability, making it difficult to respond quickly and appropriately. This resulted in issues such as longer response times and an increased risk of outrage.

[1948] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1949] In this invention, the server includes means for providing appropriate responses to complaints, means for preventing further flaming, means for shortening response time, means for displaying a reliability score for the response generated by the AI, means for reducing the variability in response skills by having an operator refer to the response generated by the AI, and means for a security operator to receive the AI's suggestions and respond while checking the reliability score. This allows the operator to respond quickly and appropriately while gaining peace of mind.

[1950] A "contact person" is a person whose job is to directly respond to inquiries and complaints from customers and users.

[1951] An "operator" is a person whose job is to operate a system or machine and perform a specific task.

[1952] A "complaint handling support system" is a system that supports appropriate responses to complaints from customers and users.

[1953] "Means for providing appropriate responses" refers to the function of generating accurate and effective responses to complaints and providing them to operators.

[1954] "Measures to prevent flame wars" is a function that takes measures to prevent problems from escalating when dealing with complaints.

[1955] "Means for shortening response time" refers to a function for reducing the time it takes to respond to complaints.

[1956] "Means for displaying a reliability score for responses generated by AI" is a function that quantifies the reliability of responses generated by AI and displays it to the operator.

[1957] "Means to reduce variability in response skills by having operators refer to responses generated by AI" is a function that reduces the differences in response skills between operators by having operators refer to responses generated by AI.

[1958] A "security operator" is a person in charge of security-related tasks and whose job is to ensure the safety of systems and facilities.

[1959] A "confidence score" is a number used to evaluate the reliability of a response generated by AI.

[1960] As an embodiment of the present invention, a complaint handling support system used by a security operator will be specifically described.

[1961] System configuration

[1962] The system consists of the following major components:

[1963] 1. Server: Hosts the AI ​​model and generates responses to queries.

[1964] 2. Terminal: A device such as a smartphone or tablet used by a security operator.

[1965] 3. User interface: An application where operators can enter queries and see the AI's responses and confidence scores.

[1966] Hardware and software used

[1967] Hardware: smartphones, tablets, servers

[1968] Software: Flask (a Python web framework), requests (a library for sending HTTP requests), generative AI models

[1969] Data processing and calculation

[1970] The server receives queries from security operators and sends prompts to the generative AI model, which generates responses to the queries and returns the responses to the server, which then calculates a confidence score for the responses and sends it to the operator's terminal.

[1971] Processing flow

[1972] 1. Entering a query: The security operator enters a query through the user interface of the terminal.

[1973] 2. Sending a query: The terminal sends a query to the server.

[1974] 3. Generate response: The server sends a query to the generative AI model and generates a response.

[1975] 4. Calculate confidence score: The server calculates a confidence score for the generated response.

[1976] 5. Displaying the response and score: The server sends the response and confidence score to the terminal for the operator to review.

[1977] Specific examples

[1978] For example, a security operator enters a query such as "What should we do if a suspicious person breaks into a building?" This query is sent to the server, and the generative AI model generates an appropriate response. The server calculates a confidence score for the generated response and displays the response and score on the operator's device.

[1979] Prompt Sentence Examples

[1980] What should you do if a suspicious person breaks into your building?

[1981] In this way, security operators can take prompt and appropriate action while referring to AI suggestions, which will provide peace of mind and reduce the variability in response skills.

[1982] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1983] Step 1:

[1984] A security operator enters a query through a terminal user interface.

[1985] Input: The query entered by the security operator (e.g., "What should we do if a suspicious person breaks into our building?")

[1986] Output: The query is entered into the terminal.

[1987] How it works: The security operator opens the application on their smartphone or tablet, enters a query in the text box, and presses the "send" button.

[1988] Step 2:

[1989] The device sends a query to the server.

[1990] Input: The query typed into the terminal

[1991] Output: A query is sent to the server.

[1992] Specific operation: The terminal application sends the entered query to the server as an HTTP request.

[1993] Step 3:

[1994] The server sends queries to the generative AI model and generates a response.

[1995] Input: The query sent to the server

[1996] Output: The response from the generative AI model

[1997] Specific operation: The server sends the received query as a prompt to the generative AI model, and the generative AI model generates a response to the query and returns it to the server.

[1998] Step 4:

[1999] The server calculates a confidence score for the generated response.

[2000] Input: Response from a generative AI model

[2001] Output: Confidence score

[2002] Specific operation: The server analyzes the content of the generated response and applies a confidence score calculation algorithm to calculate a confidence score.

[2003] Step 5:

[2004] The server sends the response and the confidence score to the device.

[2005] Input: Response and confidence score from the generative AI model

[2006] Output: Response and confidence score sent to the device

[2007] Specific operation: The server combines the response and the confidence score and sends it to the terminal as an HTTP response.

[2008] Step 6:

[2009] The terminal displays the response and confidence score to the security operator.

[2010] Input: The response and confidence score sent by the server

[2011] Output: Response and confidence score displayed to the security operator

[2012] Specific operation: The terminal application displays the received response and confidence score on the screen for the security operator to review.

[2013] Example 3

[2014] Next, a third embodiment of the third embodiment will be described. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[2015] Conventional complaint handling support systems did not provide enough support for counter staff and operators to respond appropriately, making it difficult to reduce response times, provide psychological security, and standardize response skills.In addition, it was difficult to improve the skills of operators and maintain their motivation, which led to issues with continuing employment.

[2016] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[2017] In this invention, the server includes a means for generating a response to a prompt using a generative AI model, a means for evaluating the quality of the generated response, and a means for providing feedback based on the evaluation results. This allows operators to respond appropriately and quickly, improving their skills and maintaining their motivation. It also provides a sense of security and standardizes response skills, resolving the issue of continued employment.

[2018] A "customer service representative" is an employee whose role is to directly respond to customer inquiries and complaints.

[2019] An "operator" is a person who operates the system and handles customer service and complaints.

[2020] A "complaint handling support system" is a system that provides support for appropriate responses to customer complaints.

[2021] A "generative AI model" is a model that uses artificial intelligence to generate a response to an input prompt.

[2022] A "prompt sentence" is text such as a question or instruction that a user enters into a syst...

Claims

1. A first means for receiving a query from a user; a second means for generating a first prompt sentence for outputting a first response to a received inquiry from the user; a third means for inputting the generated first prompt sentence into a generative AI model to generate the first response; a fourth means for calculating a confidence score for the generated first response; A fifth means for displaying the generated first response and the calculated confidence score on a terminal of a person in charge; A system including:

2. a sixth means for analyzing an emotional state of the user from text of a query from the user using an emotion engine; the second means generates, based on the query from the user and the emotional state of the user, a second prompt sentence for outputting a second response to the query in accordance with the emotional state of the user; The third means inputs the generated second prompt sentence into a generative AI model to generate the second response. The system of claim 1 .

3. A seventh means for predicting a risk of escalation from the inquiry from the user and the emotional state of the user. The system of claim 2.

Images