NLP-based car insurance business conversation clue recommendation method and related device thereof

Abstract

This application belongs to the field of financial technology and is applied to the field of insurance pricing guidance in financial business. It relates to a method and related equipment for recommending call leads in auto insurance business based on NLP. The method includes: recognizing recorded human-computer dialogue; performing text-to-text processing to dynamically acquire dialogue text data in real time; using NLP (Natural Language Processing) to process the dialogue text data; obtaining customer intent clustering results; recognizing the dialogue intent of the calling customer; and selecting the corresponding call lead based on the dialogue intent and assigning it to the next agent after the current human-computer dialogue, thus completing the call lead recommendation for the agent. This application uses an ASR+Kafka+dynamic data stream acquisition framework+NLP approach, enabling human customer service representatives to quickly locate the dialogue intent of customers previously served by the AI ​​voice service assistant when faced with lengthy AI intelligent outbound call data records. This improves the efficiency of human agents and ensures a good call experience for customers.

Description

Technical Field

[0001] This application relates to the field of financial technology, and in particular to a method for recommending call leads in auto insurance business based on NLP and related equipment. Background Technology

[0002] With the popularization and development of artificial intelligence and voice assistants, more and more insurance companies are using AI voice to serve users. This convenient online service saves a lot of manpower. At the same time, compared with traditional human customer service, AI voice service assistants are more controllable and have a higher degree of parallelism. However, at present, because the "AI boom" in the insurance industry is still in its early stages, current AI assistants are not enough to complete the entire sales process. Therefore, there is a practice where AI voice service assistants contact customers in advance to explore their insurance needs, and then the final insurance application step is handled by human customer service.

[0003] This new approach of combining AI voice service assistants with human customer service helps save on manpower costs while ensuring the stable operation of the insurance application process. However, when human customer service representatives transfer customers who have been served by the AI ​​voice service assistant, they often find it difficult to understand and directly utilize the lengthy AI intelligent outbound call data records. Summary of the Invention

[0004] The purpose of this application is to propose an NLP-based method for recommending call leads in auto insurance business and related equipment, in order to solve the problem in the prior art that human customer service representatives often find it difficult to understand and directly utilize the lengthy AI intelligent outbound call data records when transferring customers who have been served by AI voice service assistants.

[0005] To address the aforementioned technical problems, this application provides an NLP-based method for recommending call leads in auto insurance business, employing the following technical solution:

[0006] A method for recommending leads in auto insurance business based on NLP includes the following steps:

[0007] Based on ASR technology, the recordings of human-computer dialogues between AI voice robots and customers during calls are identified.

[0008] The human-computer dialogue recording is converted to text using ASR recognition technology, and the dialogue text data corresponding to the human-computer dialogue recording is dynamically acquired in real time.

[0009] The dialogue text data is quantized using NLP (Natural Language Processing) technology to obtain the quantization results;

[0010] The quantization results are clustered based on a pre-constructed semantic topology graph of auto insurance to obtain customer intent clustering results.

[0011] Based on the customer intent clustering results and preset filtering rules, the conversational intent of the calling customer is identified;

[0012] Based on the stated dialogue intent, the corresponding response text is selected as a recommendation clue and assigned to the next agent after the current human-computer dialogue, thus completing the recommendation of response clues for the agent.

[0013] Furthermore, the step of dynamically acquiring the dialogue text data corresponding to the human-computer dialogue recording in real time specifically includes:

[0014] The dialogue text data corresponding to the human-computer dialogue recording is obtained in real time through the Kafka message middleware.

[0015] The conversation text data within the Kafka message middleware is extracted using Apache Spark Streaming technology, where Apache Spark Streaming is a dynamic data stream batch processing framework.

[0016] Furthermore, in the process of performing the step of extracting the conversation text data within the Kafka message middleware using Apache Spark Streaming technology, the method further includes:

[0017] Based on the real-time extracted dialogue text data and preset calculation rules, calculate the batch processing interval, block interval, sliding window size, and sliding interval during extraction;

[0018] Based on the calculation results, determine whether the batch processing interval, block interval, sliding window size, and sliding interval need to be optimized.

[0019] If necessary, the batch processing interval, block interval, sliding window size, and sliding interval are optimized in parallel during the extraction process according to the preset optimization scheme.

[0020] Furthermore, the step of using NLP (Natural Language Processing) technology to quantize the dialogue text data and obtain the quantization result specifically includes:

[0021] The dialogue text data is sequentially processed using NLP (Natural Language Processing) techniques, including syntactic parsing, part-of-speech tagging, and word segmentation.

[0022] Based on the part-of-speech tagging and word segmentation results, and combined with a preset stop word form, stop words are deleted from the dialogue text data;

[0023] The word segmentation result after the stop words have been removed is used as the quantization result corresponding to the dialogue text data.

[0024] Furthermore, the step of clustering the quantization results based on the pre-constructed vehicle insurance semantic topology graph to obtain customer intent clustering results specifically includes:

[0025] The keywords in the quantification results are identified based on the semantic topology map of the vehicle insurance. The semantic topology map of the vehicle insurance is pre-set with keywords that represent different dimensions of vehicle insurance features, including the names of vehicle insurance related insurance types, insurance type codes, vehicle insurance business related departments, department codes, and commonly used business phrases.

[0026] The TF-IDF term frequency metric was used to calculate the term frequency of the keywords, and the frequency of different keywords appearing in the human-computer dialogue process of the calling customer was obtained based on the term frequency calculation results.

[0027] The different keywords are sorted according to their frequency, and the sorting result is used as the clustering result.

[0028] By referring to the preset keyword and customer intent comparison form, the customer intent clustering result corresponding to the clustering processing result is obtained.

[0029] Furthermore, the step of calculating the word frequency of the keywords using the TF-IDF term frequency metric and obtaining the frequency of different keywords appearing during the human-computer dialogue between the calling client based on the word frequency calculation results specifically includes:

[0030] Obtain the contact information of the calling customer, and based on the contact information of the calling customer, retrieve the historical dialogue text data corresponding to the calling customer from a preset dialogue text database, wherein the dialogue text database caches the quantitative processing results after each human-computer dialogue of the calling customer;

[0031] Based on the preset term frequency measurement formula: TFIDF i =F i *DF i The frequency of different keywords appearing during the human-computer interaction of the calling customer was obtained, where TF i IDF represents the frequency of keyword i in the current document. i This indicates the frequency of the reverse file corresponding to keyword i.

[0032] Furthermore, the step of identifying the conversational intent of the calling customer based on the customer intent clustering results and preset filtering rules specifically includes:

[0033] Based on preset filtering rules, the top N most frequent keywords that appear during the human-computer dialogue process of the calling customer are selected.

[0034] The customer intent clustering results corresponding to the top N keywords are obtained as the conversation intent of the calling customer.

[0035] To address the aforementioned technical problems, this application also provides an NLP-based auto insurance business call lead recommendation device, which employs the following technical solution:

[0036] A NLP-based auto insurance business call lead recommendation device includes:

[0037] The recording recognition module is used to recognize recordings of human-computer dialogue between the AI ​​voice robot and the calling customer based on ASR technology;

[0038] The text data acquisition module is used to process the human-computer dialogue recording into text using ASR recognition technology, and to dynamically acquire the dialogue text data corresponding to the human-computer dialogue recording in real time.

[0039] The quantization processing module is used to quantize the dialogue text data using NLP (Natural Language Processing) technology to obtain the quantization processing result;

[0040] The clustering processing module is used to perform clustering processing on the quantization processing results based on the pre-built vehicle insurance semantic topology map to obtain customer intent clustering results;

[0041] The dialogue intent recognition module is used to identify the dialogue intent of the calling customer based on the customer intent clustering results and preset filtering rules;

[0042] The conversation lead recommendation module is used to filter the corresponding conversation text as a recommendation lead based on the conversation intent and assign it to the next transfer agent after the current human-computer dialogue, thus completing the conversation lead recommendation for the transfer agent.

[0043] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0044] A computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the NLP-based auto insurance business call lead recommendation method described above.

[0045] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0046] A computer-readable storage medium storing computer-readable instructions, which, when executed by a processor, implement the steps of the NLP-based auto insurance business call lead recommendation method described above.

[0047] Compared with the prior art, the embodiments of this application have the following main advantages:

[0048] The NLP-based auto insurance business call lead recommendation method described in this application involves: recognizing recorded human-machine dialogue; performing text-to-text processing to dynamically acquire dialogue text data in real time; using NLP (Natural Language Processing) to process the dialogue text data; obtaining customer intent clustering results; identifying the dialogue intent of the calling customer; and selecting corresponding call text as recommendation leads based on the dialogue intent and assigning them to the next transfer agent after the current human-machine dialogue, thus completing the call lead recommendation for the transfer agent. This application employs an ASR+Kafka+dynamic data stream acquisition framework+NLP approach, enabling human customer service representatives to quickly locate the dialogue intent of customers previously served by the AI ​​voice service assistant when faced with lengthy AI intelligent outbound call data records, improving the work efficiency of human agents and ensuring a good call experience for customers. Attached Figure Description

[0049] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0051] Figure 2 A flowchart of an embodiment of the NLP-based auto insurance business call lead recommendation method according to this application;

[0052] Figure 3 yes Figure 2 A flowchart of a specific embodiment of step 203 shown;

[0053] Figure 4 yes Figure 2 A flowchart of a specific embodiment of step 204 shown;

[0054] Figure 5 yes Figure 4 A flowchart of a specific embodiment of step 402 shown;

[0055] Figure 6 A schematic diagram of the structure of an embodiment of the NLP-based auto insurance business call lead recommendation device according to this application;

[0056] Figure 7 A schematic diagram of the structure of an embodiment of the computer device according to this application. Detailed Implementation

[0057] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0058] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0059] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0060] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0061] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0062] Terminal devices 101, 102, and 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptops, and desktop computers, etc.

[0063] Server 105 can be a server that provides various services, such as a backend server that supports the pages displayed on terminal devices 101, 102, and 103.

[0064] It should be noted that the NLP-based car insurance business call lead recommendation method provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the NLP-based car insurance business call lead recommendation device is generally set in the server / terminal device.

[0065] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0066] Continue to refer to Figure 2 The diagram illustrates a flowchart of an embodiment of the NLP-based auto insurance business call lead recommendation method according to this application. The NLP-based auto insurance business call lead recommendation method includes the following steps:

[0067] Step 201: Identify the recorded human-machine dialogue between the AI ​​voice robot and the customer using ASR technology.

[0068] In this embodiment, ASR (Automatic Speech Recognition) refers to automatic speech recognition technology, which is a technology that converts human speech into text.

[0069] Step 202: Using ASR recognition technology, the recorded human-computer dialogue is converted into text, and the dialogue text data corresponding to the recorded human-computer dialogue is dynamically acquired in real time.

[0070] In this embodiment, the step of dynamically acquiring the dialogue text data corresponding to the human-computer dialogue recording in real time specifically includes: acquiring the dialogue text data corresponding to the human-computer dialogue recording in real time through the Kafka message middleware; and extracting the dialogue text data within the Kafka message middleware using Apache Spark Streaming technology, wherein Apache Spark Streaming is a dynamic data stream batch processing framework.

[0071] This architecture utilizes Kafka message middleware and the Apache Spark Streaming dynamic data stream batch processing framework to achieve real-time dynamic acquisition of conversational text data. Kafka message middleware acts as a soft cache for the conversational text data, and the Apache Spark Streaming dynamic data stream batch processing framework directly obtains the soft-cached conversational text data from Kafka message middleware. The soft cache refers to a non-database caching method, which saves database space and avoids excessive data volume in the database.

[0072] In this embodiment, during the step of extracting the dialogue text data within the Kafka message middleware using Apache Spark Streaming technology, the method further includes: calculating the batch processing interval, block interval, sliding window size, and sliding interval during extraction based on the real-time extracted dialogue text data and preset calculation rules; determining whether the batch processing interval, block interval, sliding window size, and sliding interval need to be optimized based on the calculation results; and if so, performing parallel optimization of the batch processing interval, block interval, sliding window size, and sliding interval during the extraction process according to a preset optimization scheme.

[0073] By using the Apache Spark Streaming dynamic data stream batch processing framework to directly obtain soft-cached dialogue text data from the Kafka message middleware, and combining the obtained dialogue text data, the batch processing interval, block interval, sliding window size, and sliding interval during extraction are calculated. These parameters are then optimized in parallel to ensure that a reasonable opportunity is dynamically provided for real-time acquisition of dialogue text data. This avoids acquisition delays due to excessively slow acquisition speeds or congestion due to excessively fast acquisition speeds. This approach fully leverages the performance of the Apache Spark Streaming dynamic data stream batch processing framework, ensuring the rationality of dialogue text data acquisition.

[0074] Step 203: Use NLP natural language processing technology to perform quantization processing on the dialogue text data to obtain the quantization processing result.

[0075] Continue to refer to Figure 3 , Figure 3 Yes Figure 2 Figure 3 is a flowchart of a specific embodiment of step 203 shown above, including the steps:

[0076] Step 301: Use NLP natural language processing technology to perform syntactic parsing,词性标注, and word segmentation on the dialogue text data in sequence;

[0077] Step 302: According to the results of词性标注 and word segmentation processing, and in combination with a preset stop word list, delete the stop words in the dialogue text data;

[0078] In this embodiment, the preset stop word list pre-includes common modal particles, such as "de", and common personal pronouns, such as "you", "I", "he", "we", "they", etc.

[0079] Step 303: Use the word segmentation processing result after deleting the stop words as the quantization processing result corresponding to the dialogue text data.

[0080] By performing syntactic parsing,词性标注, and word segmentation on the dialogue text data in sequence, then deleting the stop words, deleting the text data that is useless for identifying the dialogue intention, and retaining the dialogue text data with the dialogue intention, the data volume of the dialogue text data is reduced to a certain extent. When storing and analyzing in the later stage, it saves storage space and reduces the analysis complexity at the same time.

[0081] Step 204: Based on the pre-constructed semantic topology map of auto insurance, perform clustering processing on the quantization processing result to obtain the customer intention clustering result.

[0082] Continue to refer to Figure 4 , Figure 4 Yes Figure 2 Figure 4 is a flowchart of a specific embodiment of step 204 shown above, including the steps:

[0083] Step 401: Identify the keywords in the quantization processing result according to the semantic topology map of auto insurance. Among them, keywords representing different dimensions of auto insurance are preset in the semantic topology map of auto insurance, including auto insurance related insurance types, insurance type codes, auto insurance business related departments, department codes, and business common phrases;

[0084] It should be noted that the term "词性标注" in the original text seems to be a Chinese term that needs to be accurately translated according to the specific context. Here, a placeholder is used for now. You can provide the correct English translation for this part for a more accurate translation.In this embodiment, the vehicle insurance semantic topology map is pre-set with keywords representing different dimensions of vehicle insurance features, including vehicle insurance-related insurance type names, insurance type codes, vehicle insurance business-related departments, department codes, and commonly used business phrases. The vehicle insurance-related insurance type names are, for example, compulsory traffic accident liability insurance, commercial insurance, and third-party liability insurance in vehicle insurance. The insurance type codes refer to different insurance type numbers or insurance type variable names set by program developers to distinguish different insurance types. The vehicle insurance business-related departments are, for example, vehicle insurance claims department, vehicle insurance renewal department, and vehicle insurance sales department. The department codes refer to different department numbers or department variable names set by program developers to distinguish different departments. The commonly used business phrases are, for example, claims, insurance, and renewal.

[0085] Step 402: The TF-IDF term frequency metric is used to calculate the term frequency of the keywords, and the frequency of different keywords appearing in the human-computer dialogue process of the calling customer is obtained based on the term frequency calculation results.

[0086] Continue to refer to Figure 5 , Figure 5 yes Figure 4 A flowchart of a specific embodiment of step 402 shown includes the following steps:

[0087] Step 501: Obtain the contact information of the calling customer, and obtain the historical dialogue text data corresponding to the calling customer from the preset dialogue text database according to the contact information of the calling customer. The dialogue text database caches the quantitative processing results after each human-computer dialogue of the calling customer.

[0088] Step 502, according to the preset word frequency measurement formula: TFIDF i =F i *DF i The frequency of different keywords appearing during the human-computer interaction of the calling customer was obtained, where TF i IDF represents the frequency of keyword i in the current document. i This indicates the frequency of the reverse file corresponding to keyword i.

[0089] In this embodiment, Where, n i.j This represents the number of times keyword i appears in the current document j, ∑ i n i,j This represents the total number of times all keywords appear. Where |d| represents the total number of texts, including the current dialogue text and the historical dialogue text, |{:t i ∈d j} indicates that the current dialogue text and the historical dialogue text contain the keyword t. iThe total number of files is incremented by 1 to avoid the denominator being zero.

[0090] By combining historical dialogue text data of the calling customer, the frequency of different keywords appearing in all dialogue texts is obtained, thereby predicting the calling customer's dialogue intent in this call. This avoids the inability of single dialogue text data to accurately predict the customer's dialogue intent, making the prediction results more reliable and scientifically reasonable.

[0091] Step 403, and sort the different keywords according to the frequency, and use the sorting result as the clustering result;

[0092] Step 404: Refer to the preset keyword and customer intent comparison form to obtain the customer intent clustering result corresponding to the clustering processing result.

[0093] In this embodiment, a keyword-customer intent matching form is pre-set. The intent categories in the keyword-customer intent matching form include inquiry intent, agreement intent, vehicle information confirmation intent, premium information intent, insurance type related intent, expiration date inquiry intent, payment intention intent, last year's information inquiry intent, price comparison intent, and discount inquiry intent. Each intent is pre-set with one or more corresponding keywords.

[0094] By using a keyword-customer intent comparison form, it is easier to effectively identify the conversational intent of customers by combining keywords from the clustering results.

[0095] Step 205: Identify the conversational intent of the calling customer based on the customer intent clustering results and preset filtering rules.

[0096] In this embodiment, the step of identifying the conversational intent of the calling customer based on the customer intent clustering results and preset filtering rules specifically includes: filtering out the top N most frequent keywords that appear during the human-computer dialogue of the calling customer according to the preset filtering rules; and obtaining the customer intent clustering results corresponding to the top N keywords as the conversational intent of the calling customer.

[0097] Step 206: Based on the dialogue intent, select the corresponding response text as a recommendation clue and assign it to the next transfer agent after the current human-computer dialogue, thus completing the response clue recommendation for the transfer agent.

[0098] In this embodiment, after performing the step of selecting the corresponding response text as a recommended clue based on the dialogue intent and assigning it to the next transfer agent after the current human-computer dialogue, and completing the step of recommending the response clue to the transfer agent, the method further includes: providing response feedback to the caller based on the caller's contact information and the recommended response clue.

[0099] In this embodiment, the response text is a general script text pre-set according to different dialogue intentions. After the dialogue intention of the calling customer is filtered out, the corresponding response text is selected and sent to the agent who is transferred to the call, which guides the agent to carry out the response. There is no need to ask the calling customer for the dialogue intention again, saving call time and improving the agent's work efficiency.

[0100] This application achieves this by recognizing recorded human-computer dialogue; performing text-to-text processing to dynamically acquire dialogue text data in real time; employing Natural Language Processing (NLP) to process the dialogue text data; obtaining customer intent clustering results; identifying the dialogue intent of the calling customer; and selecting corresponding response text as recommendation leads based on the dialogue intent to assign to the next agent after the current human-computer dialogue, thus completing the recommendation of response leads for the transfer agent. This application uses an ASR+Kafka+dynamic data stream acquisition framework+NLP approach, enabling human customer service representatives to quickly locate the dialogue intent of customers previously served by the AI ​​voice service assistant when faced with lengthy AI intelligent outbound call data records, improving the work efficiency of human agents and ensuring a good call experience for customers.

[0101] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0102] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0103] In this embodiment, the dialogue intent of the calling customer is identified through AI intelligent speech recognition and NLP natural speech processing; the corresponding reply text is selected as a recommendation clue based on the dialogue intent and assigned to the next transfer agent after the current human-machine dialogue, thus completing the recommendation of reply clues for the transfer agent. This enables human customer service representatives to quickly locate the dialogue intent of the calling customer when transferring customers who have been served by the AI ​​voice service assistant, even when faced with lengthy AI intelligent outbound call data records, thereby improving the work efficiency of human agents and ensuring a good calling experience for customers.

[0104] Further reference Figure 6 As a response to the above Figure 2The implementation of the method shown in this application provides an embodiment of an NLP-based auto insurance business call lead recommendation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0105] like Figure 6 As shown, the NLP-based auto insurance business callback lead recommendation device 600 described in this embodiment includes: a recording recognition module 601, a text data acquisition module 602, a quantization processing module 603, a clustering processing module 604, a dialogue intent recognition module 605, and a callback lead recommendation module 606. Wherein:

[0106] The recording recognition module 601 is used to recognize recordings of human-computer dialogues between an AI voice robot and a customer using ASR technology.

[0107] The text data acquisition module 602 is used to process the human-computer dialogue recording into text using the ASR recognition result, and to dynamically acquire the dialogue text data corresponding to the human-computer dialogue recording in real time.

[0108] The quantization processing module 603 is used to quantize the dialogue text data using NLP (Natural Language Processing) technology to obtain the quantization processing result.

[0109] Clustering processing module 604 is used to perform clustering processing on the quantization processing results based on the pre-built vehicle insurance semantic topology map to obtain customer intent clustering results;

[0110] The dialogue intent recognition module 605 is used to identify the dialogue intent of the calling customer based on the customer intent clustering results and preset filtering rules.

[0111] The conversation clue recommendation module 606 is used to filter the corresponding conversation text as a recommendation clue based on the conversation intent and assign it to the next transfer agent after the current human-computer dialogue, thereby completing the conversation clue recommendation for the transfer agent.

[0112] This application achieves this by recognizing recorded human-computer dialogue; performing text-to-text processing to dynamically acquire dialogue text data in real time; employing Natural Language Processing (NLP) to process the dialogue text data; obtaining customer intent clustering results; identifying the dialogue intent of the calling customer; and selecting corresponding response text as recommendation leads based on the dialogue intent to assign to the next agent after the current human-computer dialogue, thus completing the recommendation of response leads for the transfer agent. This application uses an ASR+Kafka+dynamic data stream acquisition framework+NLP approach, enabling human customer service representatives to quickly locate the dialogue intent of customers previously served by the AI ​​voice service assistant when faced with lengthy AI intelligent outbound call data records, improving the work efficiency of human agents and ensuring a good call experience for customers.

[0113] In some specific embodiments of this example, the NLP-based car insurance business conversation lead recommendation device 600 further includes a parallel optimization module. This parallel optimization module is used during the step of the text data acquisition module 602 extracting the conversation text data from the Kafka message middleware using Apache Spark Streaming technology. Based on the real-time extracted conversation text data and preset calculation rules, the module calculates the batch processing interval, block interval, sliding window size, and sliding interval during extraction. Based on the calculation results, it determines whether optimization of the batch processing interval, block interval, sliding window size, and sliding interval is necessary. If so, it performs parallel optimization of the batch processing interval, block interval, sliding window size, and sliding interval during the extraction process according to a preset optimization scheme.

[0114] The parallel optimization module optimizes the batch processing interval, block interval, sliding window size, and sliding interval during extraction, ensuring that reasonable opportunities are dynamically provided for real-time acquisition of dialogue text data. This avoids slow acquisition speed leading to delays or excessively fast acquisition speed leading to congestion. It fully leverages the performance of the Apache Spark Streaming dynamic data stream batch processing framework, ensuring the rationality and optimal selection of dialogue text data acquisition.

[0115] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0116] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0117] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed] for details. Figure 7 , Figure 7 This is a basic structural block diagram of the computer device in this embodiment.

[0118] The computer device 7 includes a memory 7a, a processor 7b, and a network interface 7c that are interconnected via a system bus. It should be noted that only the computer device 7 with components 7a-7c is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0119] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0120] The memory 7a includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 7a may be an internal storage unit of the computer device 7, such as the hard disk or memory of the computer device 7. In other embodiments, the memory 7a may also be an external storage device of the computer device 7, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 7. Of course, the memory 7a may include both the internal storage unit and its external storage device of the computer device 7. In this embodiment, the memory 7a is typically used to store the operating system and various application software installed on the computer device 7, such as computer-readable instructions for a car insurance business call lead recommendation method based on NLP. In addition, the memory 7a can also be used to temporarily store various types of data that have been output or will be output.

[0121] In some embodiments, the processor 7b may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 7b is typically used to control the overall operation of the computer device 7. In this embodiment, the processor 7b is used to execute computer-readable instructions stored in the memory 7a or to process data, for example, to execute computer-readable instructions for the NLP-based auto insurance business call lead recommendation method.

[0122] The network interface 7c may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 7 and other electronic devices.

[0123] The computer device proposed in this embodiment belongs to the field of financial technology. This application identifies recorded human-computer dialogues; performs text-to-text processing to dynamically acquire dialogue text data in real time; uses NLP (Natural Language Processing) to process the dialogue text data; obtains customer intent clustering results; identifies the dialogue intent of the calling customer; and selects corresponding response texts as recommendation leads based on the dialogue intents, assigning them to the next agent after the current human-computer dialogue, thus completing the recommendation of response leads for the transfer agent. This application uses an ASR+Kafka+dynamic data stream acquisition framework+NLP approach, enabling human customer service representatives to quickly locate the dialogue intent of customers previously served by the AI ​​voice service assistant when faced with lengthy AI intelligent outbound call data records, improving the work efficiency of human agents and ensuring a good call experience for customers.

[0124] This application also provides another implementation, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by a processor to cause the processor to perform the steps of the NLP-based auto insurance business call lead recommendation method described above.

[0125] The computer-readable storage medium proposed in this embodiment belongs to the field of financial technology. This application identifies recorded human-computer dialogues; performs text-to-text processing to dynamically acquire dialogue text data in real time; uses NLP (Natural Language Processing) to process the dialogue text data; obtains customer intent clustering results; identifies the dialogue intent of the calling customer; and selects corresponding response texts as recommendation leads based on the dialogue intents, assigning them to the next agent after the current human-computer dialogue, thus completing the recommendation of response leads for the transfer agent. This application uses an ASR+Kafka+dynamic data stream acquisition framework+NLP approach, enabling human customer service representatives to quickly locate the dialogue intent of customers previously served by the AI ​​voice service assistant when faced with lengthy AI intelligent outbound call data records, improving the work efficiency of human agents and ensuring a good call experience for customers.

[0126] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0127] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A method for recommending call-back leads in auto insurance business based on NLP, characterized in that, Includes the following steps: Based on ASR technology, the recordings of human-computer dialogues between AI voice robots and customers during calls are identified. The human-computer dialogue recording is transcribed using ASR (Automatic Recognition) technology, and the corresponding dialogue text data is dynamically acquired in real time. This real-time dynamic acquisition of the dialogue text data includes: acquiring the dialogue text data in real time through a Kafka message middleware; after performing the step of dynamically acquiring the dialogue text data, the method further includes: extracting the dialogue text data from the Kafka message middleware using Apache Spark Streaming technology. Apache Spark Streaming is a dynamic data stream batch processing framework. In the process of extracting the dialogue text data from the Kafka message middleware using Spark Streaming technology, the method further includes: calculating the batch processing interval, block interval, sliding window size, and sliding interval during extraction based on the real-time extracted dialogue text data and preset calculation rules; determining whether the batch processing interval, block interval, sliding window size, and sliding interval need to be optimized based on the calculation results; if so, performing parallel optimization of the batch processing interval, block interval, sliding window size, and sliding interval during the extraction process according to a preset optimization scheme, thereby controlling the speed of acquiring the dialogue text data from the Kafka message middleware in real time through the parallel optimization scheme; The dialogue text data is quantized using NLP (Natural Language Processing) technology to obtain the quantization results; The quantization results are clustered based on a pre-constructed semantic topology graph of auto insurance to obtain customer intent clustering results. Based on the customer intent clustering results and preset filtering rules, the conversational intent of the calling customer is identified; Based on the stated dialogue intent, the corresponding response text is selected as a recommendation clue and assigned to the next agent after the current human-computer dialogue, thus completing the recommendation of response clues for the agent.

2. The method for recommending car insurance business call leads based on NLP according to claim 1, characterized in that, The step of quantizing the dialogue text data using NLP (Natural Language Processing) technology to obtain the quantization result specifically includes: The dialogue text data is sequentially processed using NLP (Natural Language Processing) techniques, including syntactic parsing, part-of-speech tagging, and word segmentation. Based on the part-of-speech tagging and word segmentation results, and combined with a preset stop word form, stop words are deleted from the dialogue text data; The word segmentation result after the stop words have been removed is used as the quantization result corresponding to the dialogue text data.

3. The method for recommending car insurance business call leads based on NLP according to claim 1, characterized in that, The step of clustering the quantization results based on the pre-constructed semantic topology graph of auto insurance to obtain customer intent clustering results specifically includes: The keywords in the quantification results are identified based on the semantic topology map of the vehicle insurance. The semantic topology map of the vehicle insurance is pre-set with keywords that represent different dimensions of vehicle insurance features, including the names of vehicle insurance related insurance types, insurance type codes, vehicle insurance business related departments, department codes, and commonly used business phrases. The TF-IDF term frequency metric was used to calculate the term frequency of the keywords, and the frequency of different keywords appearing in the human-computer dialogue process of the calling customer was obtained based on the term frequency calculation results. The different keywords are sorted according to their frequency, and the sorting result is used as the clustering result. By referring to the preset keyword and customer intent comparison form, the customer intent clustering result corresponding to the clustering processing result is obtained.

4. The method for recommending car insurance business call leads based on NLP according to claim 3, characterized in that, The step of calculating the word frequency of the keywords using the TF-IDF term frequency metric and obtaining the frequency of different keywords appearing during the human-computer dialogue of the calling customer based on the word frequency calculation results specifically includes: Obtain the contact information of the calling customer, and based on the contact information of the calling customer, retrieve the historical dialogue text data corresponding to the calling customer from a preset dialogue text database, wherein the dialogue text database caches the quantitative processing results after each human-computer dialogue of the calling customer; Based on the preset word frequency measurement formula: The frequency of different keywords appearing during the human-computer interaction between the caller and the client was obtained. Keywords Frequency value of occurrence in the current document Keywords The corresponding reverse file frequency.

5. The method for recommending car insurance business call leads based on NLP according to claim 4, characterized in that, The step of identifying the conversational intent of the calling customer based on the customer intent clustering results and preset filtering rules specifically includes: Based on preset filtering rules, the most frequently occurring items during the human-computer interaction process of the calling customer are selected. Key words for position; Get the top ranking The customer intent clustering results corresponding to the keywords in the position are used as the dialogue intent of the calling customer.

6. A NLP-based auto insurance business call lead recommendation device, characterized in that, The NLP-based auto insurance business call-back lead recommendation device implements the steps of the NLP-based auto insurance business call-back lead recommendation method as described in any one of claims 1 to 5, wherein the NLP-based auto insurance business call-back lead recommendation device comprises: The recording recognition module is used to recognize recordings of human-computer dialogue between the AI ​​voice robot and the calling customer based on ASR technology; The text data acquisition module is used to process the human-computer dialogue recording into text using ASR recognition technology, and to dynamically acquire the dialogue text data corresponding to the human-computer dialogue recording in real time. The real-time dynamic acquisition of the dialogue text data includes: acquiring the dialogue text data corresponding to the human-computer dialogue recording in real time through a Kafka message middleware; after performing the step of dynamically acquiring the dialogue text data corresponding to the human-computer dialogue recording, the module further includes: extracting the dialogue text data within the Kafka message middleware using Apache Spark Streaming technology, where Apache Spark Streaming is a dynamic data stream batch processing framework. The quantization processing module is used to quantize the dialogue text data using NLP (Natural Language Processing) technology to obtain the quantization processing result; The clustering processing module is used to perform clustering processing on the quantization processing results based on the pre-built vehicle insurance semantic topology map to obtain customer intent clustering results; The dialogue intent recognition module is used to identify the dialogue intent of the calling customer based on the customer intent clustering results and preset filtering rules; The conversation clue recommendation module is used to filter the corresponding conversation text as a recommendation clue based on the conversation intent and assign it to the next transfer agent after the current human-computer dialogue, thereby completing the conversation clue recommendation for the transfer agent. A parallel optimization module is used during the step of extracting the dialogue text data from the Kafka message middleware using Apache Spark Streaming technology in the text data acquisition module. Based on the real-time extracted dialogue text data and preset calculation rules, the module calculates the batch processing interval, block interval, sliding window size, and sliding interval during extraction. Based on the calculation results, it determines whether optimization of the batch processing interval, block interval, sliding window size, and sliding interval is needed. If so, according to a preset optimization scheme, the module performs parallel optimization of the batch processing interval, block interval, sliding window size, and sliding interval during extraction, thereby controlling the speed of acquiring the dialogue text data from the Kafka message middleware in real time through the parallel optimization scheme.

7. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, and the processor, when executing the computer-readable instructions, implements the steps of the NLP-based auto insurance business call lead recommendation method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the NLP-based auto insurance business call lead recommendation method as described in any one of claims 1 to 5.

Images