AI-based dynamic consultation flow design and script automatic optimization consultation system

Abstract

The present invention relates to an artificial intelligence-based dynamic consultation flow design and script automatic optimization consultation system.

Description

Technology Field

[0001] The present invention relates to an artificial intelligence-based dynamic consultation flow design and script automatic optimization consultation system. Background Technology

[0002] Recently, with the rise of e-commerce via the Internet, there is a demand for the development of e-commerce technologies and systems to implement them, related to intangible financial products such as deposits, insurance, and securities, in addition to conventional tangible goods.

[0003] In particular, regarding the subscription to financial products, it is legally essential to obtain sufficient explanation and consent; however, traditionally, this process was cumbersome and inconvenient because it was entirely conducted through paper documentation. Furthermore, when examining the process of explaining financial products (such as insurance products) to customers, explanations were mostly carried out by insurance agents visiting customers in person and simply reading out product details, or by insurance consultants unilaterally reading out product information over the phone. Since only formal consent was obtained from customers, there was a limitation in that it was difficult to convey sufficient information regarding the actual insurance product being purchased.

[0004] For instance, there may be cases where insurance agents or telephone consultants read through the details of an insurance product quickly and inaccurately, or intentionally omit certain information, due to time constraints or negligence of duty. In such instances, the customer may later raise an issue regarding the insurer's failure to fulfill its duty of explanation; consequently, this carries the potential for disputes between the customer and the financial institution (insurer) due to the risk of mis-selling.

[0005] There is a limitation in that the accuracy of the recording files varies significantly depending on the language habits, proficiency, and pronunciation / vocalization characteristics of the person explaining the product (e.g., insurance agent, telephone insurance consultant). For instance, there is a risk that the meaning conveyed to the customer may differ from the actual wording due to issues with the pronunciation, speed, and intonation of the person explaining (reading) the insurance product. Furthermore, reading a large amount of content can lead to omissions or errors, potentially causing disputes with customers when insurance claims are paid.

[0006] Furthermore, individual agents face significant difficulties in systematically classifying and backing up recording files regarding numerous customer products; consequently, there are limitations to the current method where product scripts are managed using Excel files, and agents manually select and read the script during consultations. Prior art literature

[0007] Korean Registered Patent No. 10-2687797 The problem to be solved

[0008] The objective of the present invention, aimed at solving the aforementioned problems, is to propose a technology that goes beyond the method of a counselor simply explaining verbally, provides a more convenient counseling service environment for both counselors and customers, and maximizes counseling efficiency.

[0009] The technical problems to be solved in the various embodiments are not limited to those mentioned above, and other unmentioned technical problems may be considered by those skilled in the art from the various embodiments described below. means of solving the problem

[0010] The present invention is a consultation system that supports customer consultation based on automatic script calculation, comprising: a mediation module that connects a customer terminal and a counselor terminal in a call center environment; a script generation module formed to generate a script in a preset manner according to a service or product provided to a customer; a script selection module in which at least one script generated by the script generation module is selected in a preset manner; a script playback module that converts the script selected by the script selection module into a voice file and then executes playback to the customer through the customer terminal; and a script evaluation module that evaluates the suitability of the script based on customer selection information or response information; wherein the script generation module comprises a flow design unit that designs a consultation flow through a first artificial intelligence model, and the n+1th consultation flow is generated using customer selection information or response information regarding the first to nth consultation flows. A consultation system is provided that includes a script processing unit that generates a script corresponding to an individual consultation flow, and, based on STT (Speech To Text), converts the customer's selection information or response information into text, analyzes it in a preset manner through a second artificial intelligence model, and calculates a script suitable for the selection information or response information.

[0011] The script operation unit is configured to receive the first to m-th scripts output from the second artificial intelligence model, and to enable the counselor to monitor the script provision status, and to output the k-th script selected by the counselor in the script selection module (where k is a natural number greater than or equal to 1 and less than or equal to n), and the second artificial intelligence model can be machine learned using the first to m-th scripts and the k-th script as a training dataset.

[0012] The above script generation module further includes a speech information analysis unit that extracts key keywords and contextual features from the customer's response information converted into text, and derives the customer's actual speech information based on the extracted information; wherein the speech information analysis unit calculates speech information about the customer by analyzing the customer's sentiment information, intention information, and context information using keywords of the response information based on natural language processing (NLP) in the second artificial intelligence model, and the script calculation unit may be configured to output a script corresponding to the speech information.

[0013] The script evaluation module above includes a first evaluation unit that evaluates a consultation flow generated by the flow design unit; and the first evaluation unit may be configured to calculate the flow redundancy of each of the first to p-th consultation flows when the first to p-th consultation flows proceed through customer selection information or response information, wherein each of the first to p-th consultation flows includes at least one predetermined meta value, and after calculating the flow redundancy based on the meta value, evaluate the design appropriateness of the consultation flow in a predetermined manner by considering the flow redundancy.

[0014] The script evaluation module includes a second evaluation unit that evaluates a script generated by the script operation unit using customer feedback; the second evaluation unit is configured to evaluate the appropriateness of the script by considering (a1) customer's understanding information regarding the script based on customer's selection information or response information; and (a2) customer's positive behavior information and negative behavior information after the script is played by the script playback module; and when machine learning the second artificial intelligence model, the appropriateness score of the script can be applied as a weight and processed as learning information.

[0015] The above script playback module is configured to enable bidirectional transmission of the voice of a script provided through the script processing unit during a voice call between a customer terminal and a counselor terminal by linking with an IP PBX through the above mediation module, and the above mediation module can be configured to perform Enterprise Application Integration (EAI) specialized communication between the customer terminal and the counselor terminal, provide a video call function, and allow a counselor to send a video consultation URL to a customer via SMS or message, so that when the customer clicks the URL, a video consultation is automatically connected.

[0016] In addition, the present invention provides a method using the aforementioned consultation system, comprising: (b1) connecting a customer terminal and a counselor terminal in a call center environment by means of an intermediary module; (b2) designing a consultation flow in a pre-set manner according to a service or product provided to a customer by means of a script generation module, and generating a script for said consultation flow; (b3) selecting one of the scripts generated in step (b2) by means of a script selection module and providing it to a customer; (b4) playing the script selected in step (b3) to a customer through said customer terminal by means of a script playback module; (b5) performing a consultation with a customer according to the consultation flow designed in step (b2), wherein the consultation flow or the script is modified in said script generation module based on the customer's feedback regarding the script played to the customer in step (b4); and (b6) in the script evaluation module, the suitability of the script is evaluated in a preset manner based on customer feedback obtained in step (b5), and a first artificial intelligence model that designs a consultation flow and a second artificial intelligence model that generates a script are each independently machine-learned using the evaluation results; the method is provided.

[0017] Furthermore, the present invention relates to a consultation system that supports customer consultation based on automatic script calculation. More specifically, the system includes a mediation module that supports smooth consultation by connecting a customer terminal and a counselor terminal in a call center environment, a script generation module that automatically generates a script in a preset manner according to the type and characteristics of a service or product provided to a customer, a script selection module that automatically selects at least one of a plurality of scripts generated by the script generation module according to preset criteria, a script playback module that converts the script selected by the script selection module into an audio file and plays it to the customer through the customer terminal, and a script evaluation module that evaluates the suitability of the script used during the consultation process based on the customer's selection information or response information.

[0018] The script generation module provided by the consultation system of the present invention performs the function of designing a consultation flow and includes a flow design unit that automatically designs the consultation flow using a first artificial intelligence model. The flow design unit continuously collects customer selection information and response information generated during the consultation process and optimizes the consultation flow based on this information. When generating the n+1th consultation flow, it is configured to design a more sophisticated consultation flow by learning the customer selection information and response information from the 1st to nth consultation flows collected in the previous consultation. In this way, since the consultation flow is generated dynamically by reflecting actual customer response patterns and selection information rather than using a predetermined, uniform method, it is possible to conduct consultations that are more effective and personalized than existing standardized consultation methods.

[0019] In addition, the script generation module provided by the counseling system of the present invention includes a script processing unit that performs the function of generating individual counseling scripts corresponding to the counseling flow. The script processing unit is configured to automatically generate a script suitable for the selected information or response information by converting the customer's selection information or response information from speech to text using STT (Speech To Text) technology, analyzing the text data using a second artificial intelligence model, and performing calculations in a preset manner. Accordingly, an optimal counseling script can be generated and provided in real time based on the customer's response. This has the advantage of reducing the workload of counselors and maximizing counseling efficiency compared to the method in existing counseling systems where counselors must directly lead the conversation and respond.

[0020] The script selection module provided by the consultation system of the present invention performs the role of selecting one of a plurality of consultation scripts generated by the script generation module, and the selection process is carried out according to pre-set criteria by comprehensively considering the consultation progress, the customer's past selection history, and the characteristics of the services and products currently being provided. Accordingly, even for consultations regarding the same product or service, the most suitable script can be automatically selected according to the customer's tendencies and response patterns, thereby further improving the accuracy and personalization level of the consultation.

[0021] In addition, the script playback module provided by the counseling system of the present invention converts the counseling script selected by the script selection module into an audio file and plays it back to the customer through a customer terminal. It is configured to provide counseling content in a natural and clear voice by applying Text-to-Speech (TTS) technology. Through this, customers can receive clear auditory guidance during the counseling process, and counseling can be conducted in an automated manner without direct interaction by a counselor, thereby enhancing the operational efficiency of the counseling center.

[0022] Furthermore, the script evaluation module provided by the consultation system of the present invention performs the function of evaluating the suitability of scripts provided during the consultation process based on customer selection and response information. This suitability evaluation is conducted by comprehensively considering the customer's reaction after the consultation ends, the consultation completion rate, and whether additional inquiries occur. Additionally, the evaluation results are reflected in the future consultation flow design and script generation process to enable continuous improvement of consultation quality.

[0023] The script computation unit described above is a module that performs the function of computationally determining an optimal consultation script based on customer selection and response information during the consultation process. The script computation unit is configured to automatically generate a script using a second artificial intelligence model and provide it to a counselor. More specifically, the script computation unit computes and outputs the first to m-th scripts (hereinafter referred to as "candidate scripts") based on customer response and selection information analyzed through the second artificial intelligence model, and the system is configured to allow the counselor to monitor the provision status of these candidate scripts in real time.

[0024] The candidate scripts generated by the above script operation unit can be directly verified by the counselor, and the counselor is provided with the ability to select and use a specific script (hereinafter referred to as "the k-th script") that is judged to be most appropriate for the current counseling situation. Here, k is defined as a natural number greater than or equal to 1 and less than or equal to m, and this refers to the specific script that the counselor judges to be the most suitable among the candidate scripts. That is, the counselor can actively select a script suitable for the flow of conversation with the customer from among a number of candidate scripts, thereby ensuring flexibility in counseling.

[0025] Furthermore, in the present invention, the second artificial intelligence model is configured to continuously undergo machine learning by utilizing the k-th script selected by the counselor and the first to m-th scripts generated by the script processing unit as a training dataset. More specifically, the k-th script selected by the counselor during a specific counseling process signifies that it was the most appropriate response method at that time of counseling, and this can be utilized as important training data for generating more sophisticated scripts in similar counseling situations in the future. Therefore, the counseling system of the present invention is characterized by not merely storing existing counseling data, but by reflecting the script information selected by the counselor in real time to continuously improve the counseling flow and script generation algorithm.

[0026] Furthermore, the second AI model is designed to learn the response methods preferred by the agent and customer reaction patterns in specific consultation situations by comparing and analyzing the selected script with the remaining candidate scripts not chosen by the agent. In other words, if the agent repeatedly selects a specific type of script, it indicates that the response method of that type is effective for that customer type or consultation situation; by continuously reflecting this data, the quality of the automated response of the consultation system can be improved.

[0027] The above script generation module includes various functions to improve the quality of consultation by more precisely analyzing customer response information obtained during the consultation process with the customer, and more specifically, includes a speech information analysis unit that converts the customer's response information into text, extracts key keywords and contextual features from the response information, and derives the customer's actual speech information based thereon.

[0028] The aforementioned intent information analysis unit is configured to accurately identify the customer's actual intent by analyzing the customer's response information in greater depth. To this end, it utilizes Natural Language Processing (NLP) technology to extract core keywords with specific meanings from the response information and analyzes contextual features related to those keywords to derive the essential meaning the customer intends to convey.

[0029] In particular, the aforementioned opinion information analysis unit performs the function of analyzing customer response information from various angles by utilizing a second artificial intelligence model, and specifically consists of the following process.

[0030] First, the voice responses provided by the customer during the consultation process are converted into text data using STT (Speech to Text) technology. The converted text data is used as foundational data to analyze the customer's intent.

[0031] Second, a keyword extraction process is performed to identify key keywords in the converted text data. In this process, key words with meaning within customer responses are identified using a pre-established keyword database or a deep learning-based keyword extraction algorithm.

[0032] Third, the context of customer responses is identified by analyzing the relationships between keywords and contextual characteristics. To this end, syntactic parsing and semantic analysis techniques are applied to analyze the relationships between words within a sentence, thereby more precisely inferring the emotions, intentions, and context contained in the customer responses.

[0033] Fourth, based on the keywords and contextual characteristics analyzed above, the actual intent information that the customer intends to express during the consultation process is derived. Here, intent information includes not only the content directly expressed by the customer but also indirectly indicated intentions, emotional states, and problem-solving requirements, and techniques such as sentiment analysis, intent recognition, and context prediction are applied to analyze this more accurately.

[0034] The aforementioned communication information analysis unit performs the function of inferring comprehensive communication information, including the hidden intentions and emotions of the customer's response, rather than simply converting the response into text; the communication information derived in this way is utilized as important foundational data to provide more natural and appropriate responses in the subsequent consultation flow.

[0035] In the present invention, the script processing unit is configured to generate and output an optimal script that meets the customer's requirements and consultation objectives based on the customer's opinion information derived from the opinion information analysis unit. To this end, the script processing unit possesses a number of pre-learned consultation script patterns and, after identifying the pattern corresponding to the customer's opinion information, is configured to automatically select and provide the most suitable consultation script.

[0036] For example, if a customer responds that "the fee is too expensive" during a consultation, instead of simply responding based on the word "expensive," the aforementioned sentiment analysis unit performs sentiment analysis to determine that the customer is raising a complaint, and through intent analysis, it can infer that the customer wants information regarding fee adjustments or discount benefits. Based on these analysis results, the script processing unit is configured to automatically output a response script that can alleviate the customer's complaint.

[0037] In addition, if the customer's response is ambiguous or their emotional state is negative, the script processing unit can respond by sending a signal to allow a counselor to intervene immediately, or by outputting a script containing additional verification questions to enable a more careful response.

[0038] The above script evaluation module is configured to evaluate the appropriateness of the consultation flow used during the consultation process and to continuously improve the performance of the consultation system based on this, and more specifically, includes a first evaluation unit that performs the function of quantitatively analyzing and evaluating the quality of the consultation flow generated by the consultation flow design unit.

[0039] The first evaluation unit described above is configured to utilize customer choice information or response information collected during the actual consultation process with the customer to analyze the structural appropriateness and efficiency of a specific consultation flow when it is in progress, and to evaluate the quality of the consultation flow based on this. To this end, the first evaluation unit includes a function to calculate flow redundancy for multiple consultation flows, thereby enabling the examination of redundancy in consultation flows and the design of an optimal consultation path.

[0040] The consultation flow analyzed by the first evaluation unit described above may consist of first to p-th consultation flows, and each consultation flow may include at least one predetermined meta value. Here, the meta value is utilized as an indicator representing specific characteristics of the consultation flow and may include various information, for example, such as the duration of the consultation flow, customer response patterns, frequency of major keyword appearances, consultation completion rate, and consultation dropout rate.

[0041] In the counseling system of the present invention, the first evaluation unit performs the function of calculating flow redundancy based on the meta-value of each counseling flow, and the more detailed calculation process is performed as follows.

[0042] First, during a consultation with a customer, the system records the customer's selection or response information in real time and stores the history of each consultation flow. This enables the analysis of how frequently a specific consultation flow was used and whether the customer effectively obtained the desired information through that flow.

[0043] Second, after the consultation is concluded, a process of calculating flow redundancy is performed for each of the first through the p-th consultation flows. Here, flow redundancy is an indicator representing the degree to which a specific consultation flow is structurally similar to or overlaps with other consultation flows, and it is used to evaluate whether there are unnecessarily repeated response paths in consultations with the same customer type or the same consultation objective.

[0044] Third, in the process of calculating flow redundancy, the calculation is performed based on the meta-values ​​contained in each consultation flow. The system is designed so that flow redundancy increases as the frequency with which a specific consultation flow traverses the same consultation path or repeats similar question-and-answer patterns increases compared to other consultation flows. This enables the identification of cases where unnecessary duplication occurs during the consultation process and allows for the design of more efficient consultation flows.

[0045] Fourth, considering the flow redundancy calculated above, a process is performed to evaluate the design appropriateness of the consultation flow. Here, the evaluation of design appropriateness is carried out in a pre-set manner, and the quality of the consultation flow can be quantitatively evaluated by comprehensively analyzing the extent to which the consultation flow effectively contributes to customer response, whether the consultation duration is appropriate, and the correlation with customer satisfaction.

[0046] For example, if the degree of redundancy in a specific consultation flow is excessively high, it is highly likely that the flow is repeatedly using the same or similar paths as existing consultation flows; in this case, improvements can be made by the consultation flow design department to optimize the flow or eliminate unnecessary duplication. On the other hand, if the degree of redundancy is excessively low, the consultation flow may be overly fragmented, potentially failing to effectively provide the information the customer desires; accordingly, improvements can be made by configuring a more consistent consultation flow.

[0047] In addition, the evaluation of the design appropriateness of the aforementioned consultation flow can be linked to consultation performance for each consultation flow; if a specific consultation flow has a high probability of completion or high customer satisfaction, that consultation flow receives a higher evaluation, whereas conversely, if a customer discontinues the consultation during the process or the consultation failure rate is high, that consultation flow may be deemed to require structural improvement.

[0048] The above script evaluation module is configured to evaluate the quality of a consultation script generated by a script processing unit and to continuously improve the performance of the consultation system based on this evaluation. More specifically, it includes a second evaluation unit that performs the function of quantitatively analyzing whether a script provided during the consultation process meets the customer's requirements and consultation objectives.

[0049] The second evaluation unit described above performs the function of evaluating the appropriateness of the script by utilizing various data collected during the actual consultation process of the customer. To this end, it is configured to evaluate the appropriateness of the script based on the information provided by the customer during the consultation process, after analyzing (a1) the customer's understanding information and (a2) the customer's positive and negative behavior information.

[0050] First, (a1) customer understanding information is an indicator of how accurately a customer has understood a script provided during the consultation process, and can be calculated by comprehensively analyzing the customer's response patterns, question frequency, and whether additional information is requested. More specifically, if a customer requests additional explanation or repeatedly requests the same information after the script is provided, it may be judged that the script did not sufficiently aid the customer's understanding; conversely, if the customer makes the correct choice immediately or the consultation proceeds smoothly, the script may be evaluated as effective in improving the customer's understanding. To this end, the consultation system analyzes the customer's conversation flow in real time and quantifies the customer's reaction after a specific script is provided, enabling it to be used as an understanding indicator.

[0051] In addition, (a2) positive and negative behavioral information of the customer is used as an indicator to evaluate the impact of the script provided to the customer during the consultation process, wherein positive behavioral information refers to actions indicating that the customer has successfully achieved their desired goal after the consultation (e.g., purchasing a product, applying for a service, resolving a problem, etc.), and negative behavioral information refers to cases where the customer expresses dissatisfaction or discontinues the consultation, leading to negative results. To this end, the consultation system of the present invention is configured to monitor the behavioral patterns exhibited by the customer after receiving the script in real time and to evaluate the validity of the script based on the customer's response data.

[0052] For example, if a customer immediately purchases a product or decides to use a service based on the content of a specific consultation script after it has been provided, that script can be evaluated as an effective response method. Conversely, if the customer discontinues the consultation or provides negative feedback (e.g., "it is difficult to understand," "I need more information") after the same script is provided, that script may be deemed to require improvement. To this end, the consultation system is designed to evaluate the appropriateness of the script by comprehensively analyzing not only the customer's responses but also behavioral data following the consultation.

[0053] In the counseling system of the present invention, the second evaluation unit is designed to evaluate the appropriateness of a script and then reflect the evaluation result in the machine learning process of the second artificial intelligence model. More specifically, the evaluation result of a specific script is calculated as an appropriateness score of the script, and this appropriateness score is configured to be applied as a weight when machine learning the second artificial intelligence model and processed as learning information.

[0054] In other words, the counseling system can utilize script evaluation data accumulated during the counseling process to distinguish between effective and ineffective scripts and continuously optimize the AI ​​model by reflecting this. For instance, scripts with high customer understanding or those with a high correlation to positive behavioral information may be applied with high weights and utilized more frequently in future counseling sessions, while scripts that customers repeatedly find difficult to understand or that elicit negative reactions may be assigned low weights and deemed to require improvement.

[0055] Furthermore, the aforementioned script evaluation data can be utilized to analyze differentiated patterns based not only on individual consultation scenarios but also on specific customer groups, service types, and consultation objectives, thereby enabling the more sophisticated design of customer-tailored consultation flows. For example, it is possible to identify consultation methods preferred by customers of specific age groups and effective response script patterns for specific product categories, allowing for the continuous improvement of consultation script quality by reflecting this information.

[0056] The script playback module described above performs the role of converting a consultation script provided by a script processing unit within the consultation system into voice and providing it to the customer and the counselor. More specifically, it is configured to transmit the script voice bidirectionally through an intermediary module even during a voice call between the customer terminal and the counselor terminal while the consultation is in progress. To this end, the script playback module of the present invention is linked with an IP PBX and is designed to automatically transmit the script voice according to the flow of the consultation during a real-time voice call.

[0057] The above script playback module includes not only a method in which a counselor responds directly during a consultation, but also a function that converts a consultation script generated by the script processing unit into text-to-speech (TTS) and transmits it in real time. This enables the counselor to maximize consultation efficiency by utilizing an automatic response function when necessary. Furthermore, the voice transmission method is configured to allow for both unidirectional and bidirectional transmission, and by providing a script-based auxiliary voice while the counselor converses with the customer, it ensures the accuracy and consistency of the consultation.

[0058] In the consultation system of the present invention, the script playback module is designed to dynamically play scripts according to the consultation flow, analyzes the customer's response in real time to automatically select the necessary script, and converts the script into speech so that it can be simultaneously output to the customer terminal and the counselor terminal. For example, instead of the counselor directly explaining specific service descriptions or contract details, the system automatically provides the relevant content in speech, thereby reducing the burden on the counselor and increasing the accuracy of the consultation.

[0059] Meanwhile, the intermediary module, which is linked to the aforementioned script playback module, plays a role in supporting seamless data and voice communication between the customer terminal and the agent terminal. To this end, it is configured to connect with various internal and external systems by utilizing specialized EAI (Enterprise Application Integration) communication. The intermediary module performs the function of connecting the counseling system with an external communication network, thereby enabling stable communication between the customer and the agent even in different network environments.

[0060] Furthermore, the mediation module of the present invention is designed to provide a video call function during consultation, thereby enabling a rapid video connection when a customer requests a video consultation in addition to a voice consultation. The video call function enables real-time video consultation between the customer and the consultant, and allows for more effective responses, particularly in consultation situations such as product demonstrations, fault diagnosis, and identity verification.

[0061] To this end, the counseling system of the present invention is configured so that a counselor can send a video counseling URL to a customer using SMS or a message service, and when the customer clicks the URL, a video counseling session is automatically connected without any additional operation. Here, the URL is generated through an intermediary module, and as soon as the customer clicks, the intermediary module connects the customer terminal and the counselor terminal to initiate a video call.

[0062] The intermediary module of the present invention may utilize WebRTC (Web Real-Time Communication) or other real-time video communication protocols to enable customers to access video consultations via a browser without installing a separate application, thereby enhancing customer accessibility and maximizing the convenience of the consultation. Additionally, the intermediary module is designed to maintain consultation quality by including features such as image quality optimization, network status monitoring, and voice and video synchronization during the consultation. Effects of the invention

[0063] This invention analyzes customer responses in real time to automatically generate optimal scripts, dynamically adjusts the consultation flow, and improves consultation quality through continuous learning. This is a key feature that differentiates it from existing consultation systems; it supports counselors in interacting with customers more effectively and maximizes consultation efficiency by quickly providing the information customers desire.

[0064] The effects obtainable from various embodiments are not limited to those mentioned above, and other unmentioned effects can be clearly derived and understood by a person skilled in the art based on the detailed description below. Brief explanation of the drawing

[0065] Other aspects, features, and benefits of specific preferred embodiments of the present invention, as described above, will become more apparent from the following description in conjunction with the accompanying drawings. FIG. 1 is a conceptual diagram of a counseling system according to one embodiment of the present invention. FIG. 2 shows the configuration of a counseling system according to one embodiment of the present invention. Figure 3 shows the processing steps in the flow design unit, script operation unit, and script evaluation module of a consultation system according to one embodiment of the present invention. FIG. 4 shows the process of playing a script simultaneously on a customer terminal and a counselor terminal in a counseling system according to one embodiment of the present invention. FIG. 5 shows an exemplary UI of a counseling agent implemented on a counselor terminal in a counseling system according to one embodiment of the present invention. FIG. 6 shows the process of communication between a customer terminal and a counselor terminal in a counseling system according to one embodiment of the present invention. FIG. 7 shows the configuration of a service network using a counseling system according to one embodiment of the present invention. FIG. 8 shows the state in which a counselor of a counseling system according to one embodiment of the present invention manages and plays a script. FIG. 9 shows the process of generating and playing a voice file through real-time TTS linkage in a counseling system according to one embodiment of the present invention. FIG. 10 shows a flowchart of communication proceeding in a consultation system according to one embodiment of the present invention. FIGS. 11 to 13 show an exemplary UI on the counselor side in a counseling system according to one embodiment of the present invention. FIG. 14 is a flowchart of a method using a counseling system according to one embodiment of the present invention. It should be noted that in the drawings above, similar reference numbers are used to illustrate identical or similar elements, features, and structures. Specific details for implementing the invention

[0066] The present invention is susceptible to various modifications and may have various embodiments; specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the invention to specific embodiments, and it should be understood that the invention includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention. Similar reference numerals have been used for similar components in the description of each drawing.

[0067] Terms such as first, second, A, B, etc., may be used to describe various components, but the components should not be limited by these terms. The terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of multiple related described items or any of the multiple related described items.

[0068] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0069] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "having" are intended to specify the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0070] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0071] FIG. 1 is a conceptual diagram of a consultation system according to an embodiment of the present invention. Referring to FIG. 1, the consultation system according to the present invention is configured to display a script regarding a linked product to a counselor, and the counselor can check it in real time through a counselor agent. It may be configured to manage script generation and flow, and to display the script by configuring a display UI in a pre-set manner. In addition, the present invention can perform real-time data retrieval through integration with an EAI system. This is configured to retrieve the latest product information by linking with a legacy system and reflect it in the script.

[0072] FIG. 2 shows the configuration of a counseling system according to one embodiment of the present invention. Referring to FIG. 2, the present invention includes a mediation module (110), a script generation module (120), a script selection module (130), a script playback module (140), and a script evaluation module (150).

[0073] The intermediary module (110) includes a customer terminal management unit (111) and a counselor terminal management unit (112). It is responsible for connecting the customer terminal (101) and the counselor terminal (102). When a customer requests a consultation, the intermediary module (110) processes the request and connects to the counselor. Of course, it can also be applied in the opposite case where a counselor requests a consultation from a customer. The intermediary module (110) of the present invention is designed to interact with legacy systems, including Enterprise Application Integration (EAI) linkage, and is configured to enable real-time information provision.

[0074] The customer terminal management unit (111) is configured to manage customer information using the consultation system according to the present invention, and the counselor terminal management unit (112) is configured to manage counselor information. Through this, the consultation history information of customers and counselors can be continuously updated and managed.

[0075] The script generation module (120) includes a flow design unit (121), a first artificial intelligence model (122), a script operation unit (123), a second artificial intelligence model (124), and a pseudo-information analysis unit (125), and is configured to generate a script in a preset manner according to the service or product provided to the customer. The script generation module (120) performs the function of automatically designing a consultation flow and supporting the maintenance of an optimal consultation flow. The consultation flow is a series of processes that organize the interaction between the customer and the counselor in stages, and is designed to be flexibly adjusted according to the customer's request type, response pattern, and consultation goal. The consultation flow of the present invention has a dynamic structure that is not fixed but automatically adjusted by reflecting the customer's real-time response, and can maximize the efficiency of the consultation process.

[0076] The flow design unit (121) is configured to design a consultation flow for individual consultations. The consultation flow can be designed through the first artificial intelligence model (122). Here, the flow design unit (121) collaborates with the first artificial intelligence model (122) to learn existing consultation data and customer response information, and generates a new consultation flow based on this. The n+1th consultation flow is reconstructed in an optimal manner by analyzing the customer selection information and response information accumulated from the first to nth consultation flows. The present invention describes an exemplary consultation flow, but is not limited thereto.

[0077] The consultation flow consists of three main stages, and each stage can be automatically adjusted based on the interaction between the customer and the counselor.

[0078] The first step is the initial consultation stage, whereby when a customer requests a consultation, the process of querying the customer's basic information through the mediation module (110) and defining the purpose of the consultation is carried out. The purpose of the consultation is classified into types such as new sign-up, contract change, and service problem resolution, and based on this information, support is provided to enable the counselor to perform an appropriate response.

[0079] The second stage is the consultation stage, in which questions are asked to more specifically identify the customer's requirements, and appropriate product information or service details are provided based on the customer's response. During this process, a script automatically generated by the script processing unit (123) is displayed on the counselor's terminal so that the counselor can refer to it, and the direction of the consultation is adjusted by analyzing the customer's response in real time. Since the customer may ask unexpected additional questions or make new requests during the consultation stage, the consultation flow is configured to be flexibly adjusted according to the customer's response rather than in an existing standardized form.

[0080] The third stage is a stage for guiding the customer's decision-making, a process that supports the customer in fully understanding the consultation content and making a desired choice. In this stage, the function of the opinion information analysis unit (125), which analyzes the customer's emotional information, intention information, and context information to accurately identify the choice the customer wants, plays an important role. The opinion information analysis unit (125) is configured to utilize a second artificial intelligence model (124) to extract key keywords from the customer's response and analyze them using NLP (natural language processing) techniques to infer even hidden intentions that the customer has not directly expressed.

[0081] For example, if a customer states that "insurance premiums are expensive," the system operates by determining whether this is a simple request for a price comparison or if they are actually hesitating to sign up due to exceeding their budget, and adjusting the consultation flow accordingly. In this way, by analyzing customer responses more precisely and customizing the consultation flow, the information customers want can be provided quickly and accurately.

[0082] The first artificial intelligence model (122) is configured to design a consultation flow based on the above steps. The first artificial intelligence model (122), trained through machine learning, can design a consultation flow by considering customer information that is stored in advance or verified, and information on services / products currently provided to customers. Even for the same service / product, the consultation flow may be designed differently depending on the customer. If customer information is not stored in advance, information on customers similar to the customer currently conducting the consultation may be loaded.

[0083] The consultation flow of the present invention is designed with a fluid structure that reflects the customer's reactions and emotional changes, rather than standardized responses, and is configured to automatically select the optimal response method through real-time data analysis. The system can maximize consultation efficiency by providing the optimal consultation flow without the counselor having to search for information or deliberate on response methods separately. Through this, counselors can focus on interacting with the customer, and the customer can receive a more prompt and accurate response.

[0084] The script processing unit (123) is configured to generate a script corresponding to an individual consultation flow through the second artificial intelligence model (124). Based on Speech To Text (STT), the customer's selection information or response information is converted into text, and then analyzed in a preset manner through the second artificial intelligence model (124) to generate a script suitable for the selection information or response information.

[0085] The script operation unit (123) is configured to receive the first to m-th scripts output from the second artificial intelligence model (124). The counselor is configured to monitor the script provision status and may be configured to output the k-th script selected by the counselor in the script selection module (130) (where k is a natural number greater than or equal to 1 and less than or equal to n). The second artificial intelligence model (124) may be machine learned using the first to m-th scripts and the k-th script as a training dataset.

[0086] The second artificial intelligence model (124) receives information about the consultation flow from the flow design unit (121) and generates a script corresponding to the consultation flow. Since the second artificial intelligence model (124) performs machine learning with information including customer information, different scripts may be output depending on the individual customer, even if the consultation flow is the same. As an example, the terms used in the script (excluding essential terms such as legal terms) may be used differently depending on the customer's gender, age, occupation, etc., and accordingly, the length of the script may also vary. For example, the content of the script explaining to a college student may be structured differently from the script explaining to an office worker.

[0087] The script processing unit (123) evaluates the appropriateness of the current consultation flow based on the analysis results of the second artificial intelligence model (124) while referring to the consultation flow designed by the first artificial intelligence model (122), and supports transitioning to the next consultation step as needed. For example, if a customer says they want to compare insurance products but is hesitant because a specific product is expensive, the consultation flow can be adjusted so that it deviates from simply listing product information and instead guides the customer to additional discount benefits or suggests similar, cheaper products. This is a feature that differentiates it from existing standardized consultation methods, and is designed so that the customer's response can change the consultation flow in real time. The script processing unit (123) is configured to calculate a script corresponding to the adjusted consultation flow.

[0088] The script processing unit (123) performs script optimization and has a structure in which learning occurs automatically without the intervention of a counselor. After the consultation ends, the customer's response information and the script data selected by the counselor are used as learning information for the second artificial intelligence model (124), and the script processing unit (123) analyzes this data and updates it so that it can provide a more appropriate script when similar consultation situations occur in the future. For example, when inquiries about the same product occur multiple times, if it is analyzed that a specific type of explanation received a positive customer response, the weight is adjusted so that the corresponding script is selected preferentially. Through this learning process, the script processing unit (123) can generate increasingly sophisticated consultation scripts over time and continuously improve the quality of the consultation.

[0089] The opinion information analysis unit (125) is configured to extract key keywords and contextual features from the customer's response information converted into text, and then derive the customer's actual opinion information based on the extracted information. The opinion information analysis unit (125) is configured to communicate with the second artificial intelligence model (124) and can be understood as a separate, independent component for determining the customer's actual opinion. The opinion information analysis unit (125) is configured to use keywords of the response information based on natural language processing (NLP) in the second artificial intelligence model (124). Here, by analyzing the customer's sentiment information, intention information, and context information, the opinion information regarding the customer is computed, and the script computation unit (123) is configured to output a script corresponding to the opinion information.

[0090] The opinion information analysis unit (125) does not simply convert the customer's response into text, but rather performs the function of identifying the customer's underlying intention that is not clearly expressed by comprehensively analyzing the customer's sentiment information, intention information, and context information. This is achieved by utilizing natural language processing (NLP) techniques to extract keywords from the customer's utterance, as well as by multidimensionally analyzing the sentiment, context, and intention of the response.

[0091] The opinion information analysis unit (125) communicates with the second artificial intelligence model (124) to analyze the customer's response in real time and is linked with the consultation flow design unit (121) and the script operation unit (123) so as to adjust the consultation flow based on this. Below, the operation of the consultation system of the present invention is explained through a specific example of a method for determining the customer's actual opinion.

[0092] Example 1: Determining customer price sensitivity in insurance product consultation

[0093] When a customer inquires about signing up for insurance, the counselor generally provides a script that explains the benefits of the insurance product and encourages the customer to sign up. However, if the customer's response includes the sentence "The insurance premium is more expensive than I thought," it is difficult to provide appropriate counseling by simply extracting the keyword "expensive." The articulate information analysis unit (125) of the present invention analyzes this utterance of the customer as follows.

[0094] 1) Keyword Extraction: "Insurance premium", "expensive" 2) Sentiment Analysis: Negative sentiment (-0.75) => Indicates that the customer feels burdened by the price 3) Intent Analysis: Implicit intention to know if price adjustment is possible (Price Sensitivity) 4) Context Analysis: If the customer mentioned "wanting to compare with other insurance" in the previous conversation, they are not simply complaining but are actively seeking alternative options.

[0095] Based on these analysis results, the physician information analysis unit (125) determines that the customer is a price-sensitive customer and that it is appropriate to switch to a consultation flow that offers discount benefits or recommends cheaper alternatives. Accordingly, the script operation unit (123) is configured to display to the counselor a script that emphasizes whether a discount is applicable or cost-saving effects, rather than the existing product description script. Example 2: Judging the Willingness to Resolve a Problem in a Customer Complaint Consultation

[0096] When a customer raises a service complaint during a call center consultation, some customers want an actual solution, while others call simply to express their dissatisfaction. In existing consultation systems, the types of customer complaints are classified into standardized categories, but the opinion information analysis unit (125) of the present invention can analyze the customer's complaint more deeply to determine whether an actual solution is needed or if it is merely an expression of emotion.

[0097] If a customer says, "The internet speed is too slow. It's really frustrating," this system analyzes as follows.

[0098] 1) Keyword Extraction: "Internet speed", "slow", "frustrating" 2) Sentiment Analysis: Strong negative sentiment (-0.9) => The customer is currently in a psychologically uncomfortable state 3) Intent Analysis: "Request to resolve slow speed issue" (Technical Issue) 4) Context Analysis: Check previous consultation records => History of recent rate plan change

[0099] Based on these results, the system determines that the customer is not merely raising an emotional complaint but actually desires a technical solution. Accordingly, the agent can adjust the consultation flow to connect the customer immediately with the technical team, rather than simply offering emotional comfort. Example 3: Assessing a Customer's Investment Propensity in Financial Consultation

[0100] When a customer inquires about a specific product during a consultation at a bank or securities firm, there is a difference between recognizing the customer's response simply as a request for information and responding by considering the customer's investment propensity. The opinion information analysis unit (125) of the present invention can analyze the following when a customer asks, "The stock market is too unstable these days, is this product okay?"

[0101] 1) Keyword Extraction: "Stock Market", "Anxiety", "Will It Be Okay?" 2) Sentiment Analysis: Concern (Neutral to Negative) 3) Intent Analysis: Seeking Stable Investment Products (Risk Aversion) 4) Context Analysis: Checking the customer's past investment history => Selected products with a strong recent risk-averse tendency

[0102] Based on this, the system determines that it must consider that the customer wants not just product information, but stable investment products. Accordingly, the script operation unit (123) displays a script that explains products that guarantee a stable rate of return, rather than a general product description script. Example 4: Converting a customer's vague request into a clear intention

[0103] Some customers often ask questions in a vague manner without expressing specific requirements. For example, if a customer asks, "What is the most popular insurance these days?", the existing system can only respond by simply listing popular products. However, the opinion information analysis unit (125) of the present invention can analyze as follows.

[0104] 1) Keyword Extraction: "Popular Insurance" 2) Sentiment Analysis: Neutral 3) Intent Analysis: Likely to have a specific purpose for signing up 4) Context Analysis: If the customer is in their 30s or 40s and has children => Likely to be interested in family insurance

[0105] In this case, the system recognizes that the customer is likely to want product recommendations suitable for them rather than simply information on popular products, and supports the counselor in making customized suggestions based on this. As such, the opinion information analysis unit (125) performs a key role in deriving actual opinions by comprehensively considering the customer's emotions, intentions, and context, rather than simply analyzing keywords.

[0106] The script selection module (130) includes a script storage unit (131) and a counselor selection unit (132). The script storage unit (131) may store multiple scripts provided by the script generation module (120). The counselor selection unit (132) is configured to allow selection of any one of the multiple scripts transmitted to the script storage unit (131). The script to be output to the customer may be automatically selected, but in an environment where the counselor continuously monitors, the counselor may be configured to directly select the script.

[0107] The script playback module (140) includes a voice file conversion unit (141) and a voice file transmission unit (142). Referring first to FIG. 7, after being converted into a voice file through a script media server, the voice file transmission unit (142) is configured to play the converted voice file on a customer terminal.

[0108] The script evaluation module (150) includes a first evaluation unit (151) and a second evaluation unit (152). The first evaluation unit (151) is configured to evaluate a consultation flow generated by the flow design unit (121), and the second evaluation unit (152) is configured to evaluate a script generated by the script operation unit (123) using customer feedback.

[0109] The first and second evaluation units (151, 152) will be described with further reference to FIG. 3. FIG. 3 shows the processing steps in the flow design unit, script operation unit, and script evaluation module of a consultation system according to one embodiment of the present invention.

[0110] The first evaluation unit (151) performs the role of evaluating whether the consultation flow is designed inefficiently and determines the appropriateness of the consultation flow design by calculating the Flow Redundancy Score (FRS) that occurs during the consultation process.

[0111] The first evaluation unit (151) performs the function of calculating flow redundancy based on the meta-value of each consultation flow while the first to pth consultation flows are in progress in the consultation system. The consultation flows are formed based on predefined consultation paths and customer response patterns, and a specific meta-value is assigned to each flow. These meta-values ​​may include consultation topics, response types, consultation switching conditions, etc., and flow redundancy can be measured based on them.

[0112] As an example, if the same consultation process is repeated multiple times during the process of a customer making an "insurance inquiry," the first evaluation unit (151) can detect this pattern and evaluate the appropriateness of the consultation flow design by calculating unnecessary redundancy within the consultation process. Additionally, if a specific flow is continuously repeated for the same customer type, the necessity of the flow is reviewed to support the automatic design of a more optimized consultation process.

[0113] The second evaluation unit (152) performs the function of evaluating the appropriateness of the script provided during the consultation process based on customer feedback, and the evaluation results are used as learning information for the second artificial intelligence model (124) within the script generation module (120). In existing consultation systems, the information provided by the counselor is inconsistent, or there is a lack of criteria to objectively determine whether a specific consultation script is effective. To solve these problems, the present invention is configured to analyze the customer's reaction to the script provided during the consultation and reflect this in a machine learning model to continuously improve the quality of the consultation.

[0114] Referring to FIG. 3, individual consultation flows may be assigned multiple meta-values, and flow redundancy can be determined using these meta-values. In consultation flows that proceed continuously starting from the first consultation flow, a high flow redundancy can be interpreted as unnecessary consultation flows being repeated, and thus the flow redundancy can be input back into the first artificial intelligence model (122) as training information. As an example, if the flow redundancy of the first and second consultation flows is high, the second consultation flow may be judged to be inappropriate, and thus the flow redundancy can be applied as a weight to train the first artificial intelligence model (122).

[0115] An exemplary formula for evaluating the redundancy (flow redundancy) of a consultation flow is as follows.

[0116] [Equation 1]

[0117] R(F i ): Flow F i Degree of redundancy (Redundancy Score)

[0118] S(F i ): Flow F i Similarity Score

[0119] W i : Flow F iImportance weights

[0120] p: Total number of consultation flows

[0121] [Equation 2]

[0122] [Equation 3]

[0123] (where α, : Weight adjustment coefficient (α+ =1))

[0124] Here, FAS The larger the value, the less overlap and lower the similarity, which can be interpreted as an optimized consultation flow. Each individual consultation flow is configured to automatically assign a meta-value by the first evaluation unit (151) and to determine the similarity between the meta-values. As an example, the similarity between the m-th consultation flow and the m+1-th consultation flow is performed through a similarity calculation between the meta-values, and is configured to compare and analyze each meta-value included in the m-th consultation flow with the meta-value included in the m+1-th consultation flow. The similarity between the meta-values ​​can be defined as shown in Equation 4 below.

[0125] [Equation 4]

[0126] Here, sim(M i 1 , M i 2 ) represents the similarity between individual metavalues, and can be output as a value of 1 if they match exactly, and as a value of 0 if they do not match exactly. w i can be understood as the importance weight of the corresponding meta-value. For example, assuming that major products are important w 1 = w 2 = w 4If set to =0.2, it can be calculated as follows. Since the similarity is 0.2, it can be treated as different types of consultation flows.

[0127]

[0128] In this case, the similarity between metavalues ​​can be derived through 'cosine similarity,' which measures the degree of similarity between specific words or phrases. This method measures similarity by vectorizing words using a vector space model and calculating the angle between the two vectors. The closer the cosine similarity value is to 1, the more similar the words are; conversely, the closer it is to 0, the less related the words are. In other words, cosine similarity applied sim(M i 1 , M i 2 ) Using the value of sim(F 1 , F 2 ) It can perform operations.

[0129] The scripts (S1~Sm) presented through the script processing unit (123) and the script (S3) selected by the counselor are classified as a training dataset, and the training dataset is configured to perform machine learning on the second artificial intelligence model (124).

[0130] Meanwhile, the second evaluation unit (152) performs an analysis by considering the following elements when evaluating the script: (a1) customer's understanding of the script based on customer's choice information or response information; and (a2) customer's positive behavior information and negative behavior information after the script is played by the script playback module (140).

[0131] An exemplary formula for evaluating the appropriateness of a script is as follows. Equation 5 below is a formula for evaluating customer understanding information. The higher the value for positive indicators, the greater the customer understanding information can be derived.

[0132] [Equation 5]

[0133] U(C) : Customer Understanding Score

[0134] k : Number of customer response sentences

[0135] P(C i ): Customer Response C i Positive indicators (e.g., weighting if positive words such as "good" or "understood")

[0136] N(C i ): Customer Response C i Negative indicators (e.g., weighting if they include negative words such as "I'm not sure" or "I don't understand")

[0137] Equation 6 below is a formula for analyzing customers' positive and negative behavioral information. Here, B(C) represents the customer's behavior score. It can be configured to have a value between 0 and 1, and A pos and A neg It can be calculated based on the number of actions.

[0138] A pos : Positive behaviors (e.g., positive feedback on script content, request for continued counseling)

[0139] A neg : Negative behaviors (e.g., refusal of counseling, request for cancellation, etc.)

[0140] [Equation 6]

[0141] The script appropriateness score (SAS) can be calculated using the above Equations 5 and 6. More specifically, positive and negative indicators can be distinguished based on customer choice information and response information.

[0142] As an example, if a customer's selection information relates to an expression of intent such as cancellation or termination, it can be classified as a negative indicator, while if it relates to an expression of intent such as contract conclusion or order addition, it can be classified as a positive indicator. Of course, neutral indicators may also exist. In the case of response information, this can be performed by converting the customer's voice file into text and then analyzing it. Customer feedback is classified into positive and negative behaviors, and a specific sentence can be divided into multiple pieces of feedback. In such cases, each piece of feedback is defined as either a positive or negative behavior.

[0143] Here, P(C i ) + N(C i ) It can be calculated as =1. For example, regarding the customer's expression "I understand!" P(C i ) =1, N(C i ) It can be calculated as =0. Taking the case where the customer provided 3 feedbacks as an example, 1) P(C 1 ) =1, N(C 1 ) =0, 2) P(C 2 ) =0.5, N(C 2 ) =0.5, 3) P(C 3 ) =0, N(C 3 ) If =1, U(C)It can be calculated as follows, and the level of understanding can be evaluated as 83.3%.

[0144]

[0145] [Equation 7]

[0146] (where γ,δ: weighting coefficients (γ+δ=1))

[0147] A higher SAS value indicates a more positive customer response and that the script is functioning effectively. Here, weights for machine learning training can be applied.

[0148] [Equation 8]

[0149] (Here, is the learning weight adjustment factor)

[0150] Hereinafter, communication of a consultation system according to an embodiment of the present invention will be described with reference to FIGS. 4 to 13.

[0151] FIG. 4 shows the process of simultaneously playing a script on a customer terminal and a counselor terminal in a counseling system according to an embodiment of the present invention. On the counselor terminal, a counseling agent can proceed with the explanation by displaying a script for the corresponding product selected from the integrated terminal (information) through EAI specialized communication.

[0152] Instead of the agent reading a script aloud, an audio file can be selected from the script management folder and played simultaneously on both the customer's terminal and the agent's terminal. Here, a function can be provided to request execution after a file has been pre-generated and saved via TTS to explain the selected product, and to allow the agent to play or pause the audio.

[0153] Meanwhile, an example of a customer terminal can be implemented as a computer, portable terminal, or television capable of voice calls, capable of connecting to a remote server via a network or connecting to other terminals and servers. Here, the computer includes, for example, a notebook, desktop, or laptop equipped with a web browser, and the portable terminal can include, for example, any type of handheld-based wireless communication device that ensures portability and mobility, such as a PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, or smartphone. In addition, television may include IPTV (Internet Protocol Television), Internet TV, terrestrial TV, cable TV, etc.

[0154] The counselor terminal includes a counselor AP installed in the form of a program; for the convenience of explanation, this document uses the counselor AP for the description. The counselor AP can serve as a terminal that connects a customer's voice call transmitted from an IP exchange to provide voice calls between the counselor and the customer. It can also serve as a terminal that receives and displays customer information transmitted from the IP exchange. Here, the customer information may include voice call identification information for identifying the voice call; for example, the voice call identification information may be the caller number or the IP address of the voice call.

[0155] The agent terminal is connected to an IP network and can proceed with call procedures through an IP switch using VoIP protocols for IP calls (e.g., H.323, SIP, Skinny, etc.), thereby establishing a call session. Accordingly, the agent terminal can communicate with the IP switch while either being included within it or located outside of it. In other words, the agent terminal is an IP phone capable of installing applications and can be implemented in various forms by connecting to an IP network. It is not limited to the form factor of the agent terminal.

[0156] FIG. 5 shows an exemplary UI of a counseling agent implemented on a counselor terminal in a counseling system according to an embodiment of the present invention. In the counseling agent, a script menu, a media player, and a script window can all be implemented as UIs.

[0157] FIG. 6 shows the process of communication between a customer terminal and a counselor terminal in a counseling system according to one embodiment of the present invention, and FIG. 7 shows the configuration of a service network using a counseling system according to one embodiment of the present invention.

[0158] Through the connection with a pre-established voice telephone system (IP exchange), the product guidance voice (mp3) can be provided by transmitting it from the script playback module to both the customer and the agent's voice channels while the customer and the agent are in a voice call. Here, communication can be established in a manner that connects the script media channel and the voice telephone channel. Additionally, as illustrated in FIG. 7, while the customer and the agent are in a voice call, an invitation URL such as an SMS can be sent to the customer to connect to a video call or to display and explain the script to the customer.

[0159] FIG. 8 shows the state in which a counselor of a counseling system according to an embodiment of the present invention manages and plays a script, and FIG. 9 shows the process of generating and playing a voice file through real-time TTS linkage in a counseling system according to an embodiment of the present invention. By generating and playing a real-time voice file through real-time TTS linkage instead of a pre-generated and stored MP3 voice file, the system can be configured to provide product description audio to customers and counselors. Based on the content explained to the customer via Text to Speech, the customer's voice response is analyzed using STT (Speech to Text), and the customer's understanding information can be evaluated.

[0160] FIG. 10 shows a flowchart of communication in a counseling system according to an embodiment of the present invention. It can be processed as shown in FIG. 10. It can be connected to a script integration server via an IP switch and can be configured to convert voice files using a codec (G.711, etc.).

[0161] FIGS. 11 to 13 show exemplary UIs on the counselor side in a counseling system according to an embodiment of the present invention. These are UIs displayed on a counselor terminal, and the layout, form, and function of the UI may be changed according to the designer's choice.

[0162] FIG. 14 is a flowchart of a method using a counseling system according to one embodiment of the present invention.

[0163] Referring to FIG. 14, a method using a counseling system according to one embodiment of the present invention includes steps (S11) to (S16).

[0164] Step (S11) is a step of connecting a customer terminal and a counselor terminal in a call center environment by means of an intermediary module (110).

[0166] *Step (S12) is a step in which a consultation flow is designed in a preset manner according to the service or product provided to the customer by the script generation module (120), and a script for the consultation flow is generated.

[0167] Step (S13) is a step in which one of the scripts generated in Step (S12) is selected by the script selection module (130) and provided to the customer.

[0168] Step (S14) is a step in which the script selected in step (S13) is played to the customer through the customer terminal by the script playback module (140).

[0169] Step (S15) is a step in which a consultation with a customer is performed according to the consultation flow designed in Step (S12), and the consultation flow or script is modified in the script generation module (120) based on the customer's feedback regarding the script played to the customer in Step (S14).

[0170] Step (S16) is a step in which, in the script evaluation module (150), the suitability of the script is evaluated in a preset manner based on customer feedback obtained in Step (S15), and the first artificial intelligence model (122) that designs a consultation flow and the second artificial intelligence model (124) that generates the script are each machine-learned independently using the evaluation results.

[0171] The preferred embodiments according to the present invention described above may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the computer-readable medium may be those specifically designed and configured for the present invention, or they may be those known and available to those skilled in the art of computer software.

[0172] Examples of computer-readable media may include hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions may include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The aforementioned hardware devices may be configured to operate as at least one software module to perform the operation of the present invention, and vice versa.

[0173] In addition, the above-described method or device may be implemented by combining all or part of its configuration or function, or by implementing it separately.

[0174] Although the present invention has been described above with reference to preferred embodiments, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as described in the following claims. Explanation of the symbols

[0175] 110: Intermediary Module 120: Script generation module 130: Script Selection Module 140: Script Playback Module 150: Script Evaluation Module

Claims

Claim 1 A consultation system that supports customer consultation based on automatic script calculation, comprising: a mediation module that connects a customer terminal and a counselor terminal in a call center environment; a script generation module formed to generate a script in a preset manner according to a service or product provided to a customer; a script selection module in which at least one script generated by the script generation module is selected in a preset manner; a script playback module that converts the script selected by the script selection module into a voice file and executes playback to the customer through the customer terminal; and a script evaluation module that evaluates the suitability of the script based on the customer's selection information or response information; wherein the script generation module comprises: a flow design unit that designs a consultation flow through a first artificial intelligence model, and generates the n+1th consultation flow using the customer's selection information or response information regarding the first to nth consultation flows; and a script calculation unit that generates a script corresponding to an individual consultation flow, and, based on STT (Speech To Text), converts the customer's selection information or response information into text, analyzes it in a preset manner through a second artificial intelligence model, and calculates a script suitable for the selection information or response information. The script operation unit includes the first to m-th scripts output from the second artificial intelligence model, configured so that the counselor can monitor the script provision status, and configured to output the k-th script selected by the counselor in the script selection module (where k is a natural number greater than or equal to 1 and less than or equal to n), machine learning is performed on the second artificial intelligence model using the first to m-th scripts and the k-th script as a training dataset, and the script generation module includes a speech information analysis unit that extracts key keywords and contextual features from the customer's response information converted into text, and then derives the customer's actual speech information based on the extracted information;The method further includes, wherein the pseudo-information analysis unit computes pseudo-information about the customer by analyzing the customer’s sentiment information, intention information, and context information using keywords of the response information based on natural language processing (NLP) in the second artificial intelligence model, and the script computation unit outputs a script corresponding to the pseudo-information, and the script evaluation module includes a first evaluation unit that evaluates a consultation flow generated by the flow design unit; wherein the first evaluation unit computes the flow redundancy of each of the first to p-th consultation flows when the first to p-th consultation flows proceed through the customer’s selection information or response information, wherein each of the first to p-th consultation flows includes at least one predetermined meta-value, and after computing the flow redundancy based on the meta-value, evaluates the design appropriateness of the consultation flow in a preset manner considering the flow redundancy, and the script evaluation module includes a second evaluation unit that evaluates the script generated by the script computation unit using customer feedback; and the second evaluation unit includes (a1) the customer’s understanding information regarding the script based on the customer’s selection information or response information; and (a2) a counseling system configured to evaluate the appropriateness of the script by considering the customer's positive behavior information and negative behavior information after the script is played by the script playback module, and configured to process the appropriateness score of the script as learning information by applying it as a weight when machine learning the second artificial intelligence model.;

Images