Intelligent question and answer processing method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence and natural language processing, and discloses an intelligent question and answer processing method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a to-be-solved claim settlement problem, and calling a large model interface to perform intention recognition on the to-be-solved claim settlement problem; matching the user intention with each user role intention of a script in the dialogue script set to obtain a plurality of matching degrees; judging whether an optimal matching intention exists or not according to a preset screening condition; if the best matching intention exists, generating a guiding reply text or an answer reply text by calling a large model interface according to the best matching intention, the corresponding association knowledge table link and the claim settlement person role, and returning to the first step; and if not, calling the large model interface to generate a reply text according to the question and answer table or the knowledge table and the user intention. The customer service quality and efficiency can be comprehensively improved by combining large model intention understanding, dialogue script guidance and the claim settlement knowledge base.

Description

Technical Field

[0001] The present disclosure relates to the fields of artificial intelligence technology and natural language processing technology, and in particular to an intelligent question and answer processing method, device, electronic device and storage medium. Background Art

[0002] In modern enterprises, online customer service system is an important key channel for interaction with customers. It can respond to customer needs in a timely manner and provide customers with high-quality services, which is an indispensable part of enterprise operations.

[0003] Currently, manual online customer service is the main way for many companies to serve their customers. Although this method can provide personalized services, it faces problems such as high labor costs, slow response speed, and inconsistent service quality. Summary of the invention

[0004] In view of the above situation, the embodiments of the present disclosure provide an intelligent question and answer processing method, device, electronic device and storage medium, aiming to solve the problems of slow response speed and inconsistent service quality existing in the related technology.

[0005] In a first aspect, an embodiment of the present disclosure provides a news recommendation intelligent question-answering processing method, the method is implemented based on a claims knowledge base, the claims knowledge base includes a question-answering table, a knowledge table, and a dialogue script set, the knowledge table includes a policy information table, a claims process table, a product knowledge table, and a regulations and policies table; the method includes:

[0006] Obtain the pending claims sent by the user, and call the big model interface to perform intent recognition on the pending claims to obtain the user's intent;

[0007] The user intention is matched with the intention of each user role in the script of the dialogue script collection to obtain multiple matching degrees; and according to the preset screening conditions and the multiple matching degrees, it is judged whether there is a best matching intention to obtain a first judgment result; wherein the content of the script in the dialogue script collection includes dialogue roles, role intentions and related knowledge table links;

[0008] If the first judgment result is yes, then by calling the big model interface, according to the best matching intention, the corresponding associated knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and the process returns to the step of obtaining the claims problem to be solved sent by the user;

[0009] If the first judgment result is no, the large model interface is called to generate a reply text based on the question and answer table or the knowledge table and the user's intention and return it to the user.

[0010] In a second aspect, the embodiment of the present disclosure further provides an intelligent question-answer processing device, which is implemented based on a claims knowledge base, wherein the claims knowledge base includes a question-answer table, a knowledge table, and a dialogue script set, wherein the knowledge table includes a policy information table, a claims process table, a product knowledge table, and a regulations and policies table; the device includes:

[0011] The intention recognition module is used to obtain the pending claims sent by the user, and call the large model interface to perform intention recognition on the pending claims to obtain the user's intention;

[0012] A reply module is used to match the user intention with the intention of each user role in the script in the dialogue script set, respectively, to obtain multiple matching degrees; and based on the preset screening conditions and the multiple matching degrees, determine whether there is a best matching intention to obtain a first judgment result; wherein the content of the script in the dialogue script set includes dialogue roles, role intentions and related knowledge table links; if the first judgment result is yes, then by calling the big model interface, according to the best matching intention, the corresponding related knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and the step of obtaining the unresolved claims problem sent by the user is returned; if the first judgment result is no, then the big model interface is called, and a reply text is generated and returned to the user according to the question and answer table or the knowledge table, and the user intention.

[0013] In a third aspect, an embodiment of the present disclosure further provides an electronic device, comprising: a processor; and a memory arranged to store computer executable instructions, wherein when the executable instructions are executed, the processor executes the steps of the above-mentioned intelligent question and answer processing method.

[0014] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes the steps of the above-mentioned intelligent question and answer processing method.

[0015] By means of the above technical scheme, the intelligent question-answering processing method, device, electronic device and storage medium provided by the embodiment of the present disclosure operate based on a claim knowledge base including a question-answering table, a knowledge table (including a policy information table, a claim process table, a product knowledge table, and a regulatory policy table) and a dialogue script set. First, the unresolved claim problem sent by the user is obtained, and the large model interface is called for intent recognition. With the powerful natural language processing capability of the large model, the user's intention is deeply understood, and the subtle differences are accurately captured, laying the foundation for subsequent accurate matching of user needs. Then, with the help of the dialogue script, a clear and organized communication framework is built for the user through the predefined dialogue process, and the user is guided to have a structured dialogue, ensuring that the service is consistent from the beginning, and can be flexibly adjusted according to the user's specific questions and the dialogue context in which it is located, providing guidance that fits the user's personalized needs, which not only effectively reduces the communication confusion caused by the lack of guidance, but also reduces the dependence on manual customer service and the corresponding training costs. When the dialogue script cannot handle the user's problem, the large model is used to retrieve relevant information from the question-answering table or knowledge table, and a reply text is generated based on the relevant information. The accuracy and stability of question and answer tables or knowledge tables are crucial to reducing human errors and improving service reliability. By closely combining and coordinating the three elements of big model intent understanding, dialogue script guidance, and claims knowledge base, this solution has significant advantages over traditional online customer service solutions. It can reduce labor costs, improve response speed, ensure information reliability, enhance user experience, and guide smooth communication. It performs well in multiple dimensions and can comprehensively improve customer service quality and efficiency, help companies save resource costs and enhance competitiveness, and can also improve customer retention and sales conversion rates through personalized recommendations and care, creating greater commercial value, meeting user claims needs, and consolidating the company's service advantage.

[0016] The above description is only an overview of the technical solution of the present disclosure. In order to more clearly understand the technical means of the present disclosure, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present disclosure more obvious and easy to understand, the specific implementation methods of the present disclosure are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0018] Figure 1 A schematic diagram of an application environment of the intelligent question-answering processing method provided by an embodiment of the present disclosure is shown;

[0019] Figure 2 A schematic diagram showing a flow chart of an intelligent question-answering processing method provided by an embodiment of the present disclosure is shown;

[0020] Figure 3 A schematic diagram showing the structure of an intelligent question-answering processing device provided in an embodiment of the present disclosure is shown;

[0021] Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the technical solutions of the present disclosure will be clearly and completely described below in combination with the specific embodiments of the present disclosure and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure.

[0023] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that such usage is interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the term "including" and its variants are to be interpreted as open-ended terms meaning "including but not limited to".

[0025] As mentioned above, although the existing manual online customer service can provide personalized services, it faces problems such as high labor costs, slow response speed, inconsistent service quality, etc. Based on this, the present invention proposes an intelligent question and answer processing method, device, electronic device and storage medium, and the present disclosure is described in detail through specific embodiments below.

[0026] To facilitate understanding of this embodiment, the intelligent question-answering processing method disclosed in the embodiment of the present disclosure is first introduced in detail. The intelligent question-answering processing method provided by the embodiment of the present disclosure can be applied to Figure 1In an application environment, a client communicates with a server through a network. The server can receive the unresolved claims problem sent by the user through the client, and call the big model interface to perform intent recognition on the unresolved claims problem to obtain the user's intent; match the user's intent with the intention of each user role in the script in the dialogue script set to obtain multiple matching degrees; and judge whether there is a best matching intention based on the preset screening conditions and the multiple matching degrees to obtain a first judgment result; wherein the content of the script in the dialogue script set includes dialogue roles, role intentions and related knowledge table links; if the first judgment result is yes, then by calling the big model interface, according to the best matching intention, the corresponding related knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and the step of obtaining the unresolved claims problem sent by the user is returned; if the first judgment result is no, then calling the big model interface, according to the question and answer table or the knowledge table, and the user's intent, a reply text is generated and returned to the client, and the present invention relies on the claims knowledge base including the question and answer table, the knowledge table (including the policy information table, the claims process table, the product knowledge table, the regulations and policies table) and the dialogue script set to operate. First, the unresolved claims sent by the user are obtained, and the big model interface is called for intent recognition. With the powerful natural language processing capabilities of the big model, the user's intention is deeply understood, and the subtle differences are accurately captured, laying the foundation for subsequent accurate matching of user needs. Then, the dialogue script is used first to build a clear and organized communication framework for the user through the predefined dialogue process, guiding the user to have a structured dialogue, ensuring that the service is consistent from the beginning, and can be flexibly adjusted according to the user's specific questions and the dialogue context, providing guidance that fits the user's personalized needs, which not only effectively reduces the communication confusion caused by the lack of guidance, but also reduces the dependence on manual customer service and the corresponding training costs. When the dialogue script cannot handle the user's problem, the big model is used to retrieve relevant information from the question and answer table or knowledge table, and generate a reply text based on the relevant information. The accuracy and stability of the question and answer table or knowledge table are crucial to reducing human errors and improving service reliability. By closely combining and coordinating the three elements of big model intention understanding, dialogue script guidance, and claims knowledge base, this solution has significant advantages over traditional online customer service solutions, can reduce labor costs, improve response speed, ensure information reliability, improve user experience, guide smooth communication, and perform well in multiple dimensions, which can comprehensively improve customer service quality and efficiency. Among them, the client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices.The server is generally a computer device with a certain computing capability, which includes, for example, a terminal device or a server or other processing device, wherein the terminal device may be a user equipment (UE), a mobile device, a user terminal, a terminal, a personal digital assistant (PDA), a handheld device, a computing device, etc. The server may be implemented as an independent server or a server cluster consisting of multiple servers. In some possible implementations, the method may be implemented by a processor calling a computer-readable instruction stored in a memory. The present invention is described in detail below through specific embodiments.

[0027] Figure 2 The flowchart of the intelligent question-answering processing method provided by the embodiment of the present disclosure is shown. Figure 2 It can be seen that the embodiment of the present disclosure at least includes steps S201-S204:

[0028] S201: Obtain the unresolved claims issue sent by the user, and call the large model interface to perform intent recognition on the unresolved claims issue to obtain the user's intent.

[0029] It should be noted that the intelligent question-answering processing method provided by the embodiment of the present disclosure is implemented based on a claims knowledge base, and the claims knowledge base includes a question-answering table, a knowledge table, and a dialogue script set. The knowledge table includes a policy information table, a claims process table, a product knowledge table, and a regulations and policies table. Specifically, with respect to the knowledge table, the policy information table includes but is not limited to the policy unique identifier, the customer unique identifier, the product name, the insured amount, the premium amount, the policy effective date, the policy expiration date, the policy status, etc. The claims process table includes but is not limited to the claims process unique identifier, the insurance product name, the process step sequence number, the process step description, the required materials list, the service response time, etc. The product knowledge table includes but is not limited to the product unique identifier, the product name, the coverage description, the exemption clause, the target customer group, the premium rate, etc. The regulations and policies table includes but is not limited to the policy unique identifier, the policy name, the policy summary, the scope of application, the effective date, the expiration date, etc.

[0030] The present disclosure does not limit the method for obtaining the pending claims problem. For example, a user can enter the pending claims problem in a web form, a search box, or an input box in a mobile application, and then send it to the server deployed with the method provided by the embodiment of the present disclosure through a network request. The user can also use an intelligent voice device, such as a smart speaker, a voice assistant, etc., to input the pending claims problem by voice, and the voice data is converted into text data and sent to the server.

[0031] The large model may be, for example, ChatGPT, Doubao, Wenxiaoyan, etc., which is not limited in this embodiment. During implementation, the text of the claim problem to be solved may be cleaned up first, including removing noise words, standardization, word segmentation, etc., to improve the quality of the claim problem to be solved, thereby improving the accuracy and efficiency of the intention recognition of the large model.

[0032] The intelligent question-answering processing method provided by the present disclosure can be used in an online customer service system in the insurance field. The online customer service system can be implemented through a server, which can receive user questions in real time. For example, the claim problem to be solved is "My car was flooded in last week's rainstorm, and now I need to apply for a claim. What materials do I need to provide?" After receiving the claim problem to be solved, the online customer service system can use a large model to identify the intent of the claim problem to be solved and obtain the user's intention: the user wants to understand the claim process and the required materials.

[0033] S202: Match the user intention with the intention of each user role in the scripts in the dialogue script collection to obtain multiple matching degrees; and determine whether there is a best matching intention based on preset screening conditions and the multiple matching degrees to obtain a first judgment result; wherein the content of the scripts in the dialogue script collection includes dialogue roles, role intentions and related knowledge table links.

[0034] Each dialogue script in the dialogue script set contains possible dialogue roles, role intentions, and links to related knowledge tables. For example, the dialogue script is: "User: Greeting at first contact; System: Query treatment status & query documents to be uploaded; System: User treatment is over & the user has some documents not submitted; Customer service: Inform the user which documents to submit (link to the claim process table); User: Send document pictures; System: Upload receipts and check documents to be uploaded; System: User documents have all been submitted; Customer service: Inform the user that the documents are complete."

[0035] This embodiment does not limit the calculation method of the matching degree. During specific implementation, for example, a pre-trained language model, such as a model of the Transformer architecture, can be used to encode the user intent and the user role intent, and then the similarity score between the two is calculated to obtain the matching degree. Then, based on the preset filtering conditions and multiple matching degrees, it is determined whether there is a best matching intention to obtain a first judgment result. Here, the preset filtering conditions can be set according to actual needs. For example, if the maximum matching degree is greater than or equal to 0.9, it is determined that there is a best matching intention, and the best matching intention is the user role intention corresponding to the maximum matching degree; if the maximum matching degree is less than 0.9, it is determined that there is no best matching intention.

[0036] S203: If the first judgment result is yes, then by calling the large model interface, according to the best matching intention, the corresponding associated knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and the process returns to the step of obtaining the unresolved claims issue sent by the user.

[0037] Specifically, if there is a best matching intention, the big model can be used in combination with the preset answering method and language style of the claims adjuster role, and based on the best matching intention and the table content indicated by the corresponding associated knowledge table link, a guiding reply text or answer reply text that conforms to the logic and context can be generated and returned to the user. If the user continues to ask questions, the intent recognition, user role intent matching and other processing processes will continue to be executed.

[0038] S204: If the first judgment result is no, the large model interface is called to generate a reply text according to the question and answer table or the knowledge table and the user intention and return it to the user.

[0039] If there is no best matching intent, the large model is used to search the question and answer table or knowledge table according to the user's intent, and a reply text is generated based on the search results and returned to the user.

[0040] from Figure 2As can be seen from the method shown, this solution operates based on a claims knowledge base that includes a question-and-answer table, a knowledge table (including a policy information table, a claims process table, a product knowledge table, and a regulations and policy table) and a dialogue script set. First, the unresolved claims questions sent by the user are obtained, and the big model interface is called for intent recognition. With the powerful natural language processing capabilities of the big model, the user's intention is deeply understood, and the subtle differences are accurately captured, laying the foundation for subsequent accurate matching of user needs. Then, with the help of dialogue scripts, a clear and organized communication framework is built for users through predefined dialogue processes to guide users to have structured dialogues, ensuring that the service is consistent from the beginning, and can be flexibly adjusted according to the user's specific questions and the dialogue context in which they are located, providing guidance that fits the user's personalized needs, which not only effectively reduces the communication confusion caused by the lack of guidance, but also reduces the dependence on manual customer service and the corresponding training costs. When the dialogue script cannot handle the user's problem, the big model is used to retrieve relevant information from the question-and-answer table or knowledge table, and a reply text is generated based on the relevant information. The accuracy and stability of the question-and-answer table or knowledge table are crucial to reducing human errors and improving service reliability. By closely combining and coordinating the three elements of big model intention understanding, dialogue script guidance, and claims knowledge base, this solution has significant advantages over traditional online customer service solutions. It can reduce labor costs, improve response speed, ensure information reliability, enhance user experience, and guide smooth communication. It performs well in multiple dimensions and can comprehensively improve customer service quality and efficiency, helping companies save resource costs and enhance competitiveness. It can also improve customer retention rate and sales conversion rate through personalized recommendations and care, create greater business value, meet user claims needs, and consolidate the company's service advantage.

[0041] Furthermore, in order to better illustrate the process of the above-mentioned intelligent question and answer processing method, as a refinement and extension of the above-mentioned embodiment, the embodiment of the present invention provides several embodiments, but is not limited thereto.

[0042] In a possible implementation, the user intention is matched with the intention of each user role in the scripts in the dialogue script set to obtain multiple matching degrees, including: obtaining insurance type information involved in the claim issue to be resolved; screening the scripts in the dialogue script set according to the insurance type information to obtain a target dialogue script set; and calculating the matching degree between the user intention and the intention of each user role in each script in the target dialogue script set to obtain the multiple matching degrees.

[0043] In this embodiment, the dialogue script set is first screened according to the insurance type information involved in the claim problem to be solved to obtain the target dialogue script set. Here, the insurance type information indicates the specific insurance classification. For example, the insurance type information can be "auto insurance", "life insurance", "health insurance", etc. This embodiment does not limit the way to obtain the insurance type information. For example, when a user communicates with an online customer service system deployed with the disclosed intelligent question-and-answer method, he can send a link to the claim case he wants to consult, and the claim case link contains insurance type information. Different insurance types correspond to different dialogue script sets. For example, the dialogue script set corresponding to the insurance type of "auto insurance" may focus on the dialogue content of vehicle accident handling, compensation standards, etc.; while the dialogue script set corresponding to "health insurance" may involve more dialogue content of medical expense reimbursement, disease identification, etc. After obtaining the target dialogue script set, the matching degree between the user intention and the intention of each user role in each script in the target dialogue script set is calculated respectively, and multiple matching degrees are obtained. According to the preset screening conditions and multiple matching degrees, it is judged whether there is the best matching intention to obtain the first judgment result.

[0044] This embodiment can effectively narrow the matching scope and focus on the script content related to a specific insurance type, avoiding interference from irrelevant scripts, making intent matching more accurate and efficient, reducing unnecessary consumption of computing resources, and thereby improving the efficiency of the entire claims processing process.

[0045] In some embodiments, the dialogue script set is generated according to the following method: obtaining multiple groups of historical dialogues between users and claims adjusters; for each group of target dialogues in the multiple groups of historical dialogues between users and claims adjusters, using a large model to perform intent recognition on the dialogue text in the target dialogue to obtain a target dialogue script corresponding to the target dialogue; performing duplication elimination processing on any unprocessed dialogue script in each of the target dialogue scripts to obtain the dialogue script set; wherein, performing duplication elimination processing on the unprocessed dialogue scripts includes: for any target user intention in the unprocessed dialogue script, respectively calculating the similarity between the target user intention and each user intention in other target dialogue scripts to obtain a target similarity; if the target similarity is greater than a preset similarity, deleting the target user intention and the corresponding claims adjuster response intention, or deleting the corresponding user intention and claims adjuster response intention in other target dialogue scripts.

[0046] For example, suppose there are multiple groups of historical conversations between users and claims adjusters:

[0047] Dialogue 1:

[0048] User: "I'm sick and hospitalized. What expenses can my medical insurance cover?"

[0049] Claims adjuster: "Medical insurance generally reports hospitalization expenses, drug expenses, etc. within the scope of medical insurance. Remember to keep the invoice."

[0050] Dialogue 2:

[0051] User: "I fell down accidentally, can the accident insurance compensate me?"

[0052] Claims adjuster: "It depends on the contract. If you meet the conditions and have proof and documents, you can get compensation."

[0053] Dialogue 3:

[0054] User: "Hospitalization costs a lot of money, how much can the medical insurance cover?"

[0055] Claims adjuster: "Check the reimbursement ratio in the contract and prepare a list of expenses."

[0056] Then, for each target dialogue in multiple groups of historical dialogues between users and claims adjusters, the large model is used to perform intent recognition on the dialogue text between the user and the claims adjuster in the target dialogue, and the target dialogue script corresponding to the target dialogue is obtained:

[0057] Dialogue 1 script:

[0058] User: Inquire about the scope of medical insurance reimbursement.

[0059] Claims adjuster: Explain the scope of medical insurance reimbursement and remind you to keep the invoice.

[0060] Dialogue 2 script:

[0061] User: I would like to know whether I can claim compensation for an accidental fall under my accident insurance.

[0062] Claims adjuster: Guide the client to review the contract and inform them that they can claim if they meet the requirements.

[0063] Dialogue 3 script:

[0064] User intention: to inquire about the medical insurance reimbursement amount.

[0065] Claim adjuster's intention: Guide users to review the contract reimbursement ratio to clarify the calculation method, and remind users to prepare a list of expenses to ensure a smooth claims process.

[0066] Perform duplicate elimination processing on any pending dialogue script in each target dialogue script to obtain a dialogue script set. Assume that the dialogue 1 script is the pending script, and the preset similarity is set to 0.8. Compare the user intent of the dialogue 2 script, and calculate the similarity to be, for example, 0.2, and retain the relevant intent content of the dialogue 1 script and the dialogue 2 script; compare the user intent of the dialogue 3 script, and calculate the similarity to be, for example, 0.9. If it is greater than the preset value, then delete the user intent and the claims adjuster intent (i.e., the reply intent) in the dialogue 3 script or the dialogue 1 script. After processing all the scripts in this way, the remaining ones constitute the dialogue script set.

[0067] This embodiment obtains multiple groups of historical conversations between users and claims adjusters, uses a large model to identify the intent of the conversation text in each group of target conversations, and generates corresponding target conversation scripts based on the intent of each sentence. It then performs repetitive elimination processing to obtain a conversation script collection. This can fully mine the effective information in past real conversations, extract representative and typical conversation scripts, avoid repeated and redundant content, and make the conversation script collection more refined, accurate, and in line with actual business scenarios.

[0068] In some embodiments, a large model interface is called to generate a reply text based on the question and answer form and the user intention, including: calculating the degree of match between the user intention and each question in the question and answer form; for the largest preset number of matching degrees to be screened among each of the matching degrees, using a preset difference judgment condition to judge whether the difference between the preset number of matching degrees to be screened is large, and obtain a second judgment result; if the second judgment result is no, using the large model to select the best matching question from the questions corresponding to the preset number of matching degrees to be screened; if the second judgment result is yes, using the question corresponding to the largest matching degree to be screened as the best matching question; and using the answer to the best matching question as the reply text.

[0069] In this embodiment, the degree of match between the user's intention and each question in the question-and-answer table can be calculated first to obtain the degree of match 1, degree of match 2, ... degree of match n. Wherein n is the total number of questions in the question-and-answer table. This embodiment does not limit the method for calculating the degree of match. In specific implementation, for example, the degree of match can be calculated by counting the frequency and weight of the keywords in the user's intention in the question-and-answer table. It is also possible to use a pre-trained language model, such as a model with a Transformer architecture, to encode the user's intention and the question-and-answer table, and then calculate the similarity score between the two to obtain the degree of match.

[0070] Then, based on each matching degree, the best matching problem is screened out. Specifically, a preset number of matching degrees with the largest value can be selected from each matching degree as the matching degree to be screened. The preset number here can be set according to actual needs, and this embodiment is not limited to this. For example, it can be set to 5. Then, the preset difference judgment condition is used to determine whether the difference between these matching degrees to be screened is large. Here, the difference judgment condition can be set according to actual needs. For example, the difference judgment condition can be: calculate the variance or standard deviation of each matching degree to be screened. If the variance or standard deviation is less than the preset threshold, it means that the degree of discreteness between the matching degrees to be screened is low, and it is judged that the difference between these matching degrees to be screened is not large; if the variance or standard deviation is greater than or equal to the preset threshold, it means that the degree of discreteness between the matching degrees to be screened is high, and it is judged that the difference between these matching degrees to be screened is large. The difference judgment condition can also be: taking the maximum matching degree to be screened as the benchmark, calculating the ratio of other matching degrees to be screened to the maximum matching degree to be screened; if these ratios are within a certain range (e.g. 80%-100%), it is judged that the differences between these matching degrees to be screened are not large; otherwise, it is judged that the differences between these matching degrees to be screened are large.

[0071] After obtaining the second judgment result, if the second judgment result is that the difference between the to-be-screened matching degrees is not large, the large model is used to select the best matching question from the questions corresponding to the preset number of to-be-screened matching degrees. Here, the large model can be Doubao, Wen Xiaoyan, etc., and there is no limitation on this; if the second judgment result is that the difference between the to-be-screened matching degrees is large, the question corresponding to the maximum to-be-screened matching degree is used as the best matching question. Finally, the corresponding answer to the best matching question is returned to the user as the reply text of the to-be-resolved claim problem.

[0072] This embodiment uses the preset difference judgment condition to determine whether the difference between the largest multiple matching degrees to be screened is large. If the difference is large, it means that the question corresponding to the maximum matching degree is the best match for the question raised by the user, and it is directly used as the best matching question; if the difference is not large, the big model is used to further select the best matching question from the questions corresponding to the matching degrees to be screened, giving full play to the ability of the big model in processing complex semantics and subtle differences, ensuring that the selected best matching question is more accurate. It can be seen that this embodiment can improve the accuracy of question matching and the quality of responses as a whole.

[0073] In some embodiments, the method further includes: obtaining multiple unresolved claims issues for which intent recognition fails; respectively labeling the multiple unresolved claims issues for which intent recognition fails with correct intent to obtain a fine-tuning sample set; using the fine-tuning sample set to fine-tune the large model, and replacing the large model with the fine-tuned large model.

[0074] In specific implementation, the collected claims that failed to be recognized and have been labeled with correct intent can be cleaned first, and redundant spaces, special symbols, and format errors can be removed to standardize the content. Then, the training set, validation set, and test set can be divided according to a certain ratio (such as 8:1:1, etc.); then, the appropriate fine-tuning strategy is determined according to the large model architecture and application scenario, and hyperparameters such as learning rate, number of training rounds, and batch size are set; then, the fine-tuning sample set text and annotation information are input into the model in batches for forward propagation prediction of intent, and the loss function is used to calculate the difference between the prediction result and the actual annotation to obtain the loss value, and then the model parameters are updated according to the learning rate through the back-propagation algorithm, and the operation is repeated until the stopping condition is met; then, in the model evaluation and optimization stage, the validation set is used to evaluate the model regularly during training, and the performance is judged according to relevant evaluation indicators. If overfitting occurs, it is optimized and adjusted. After the training, the test set is used for comprehensive evaluation. If the performance is not ideal, the training is adjusted again; finally, in the model deployment stage, the fine-tuned model that meets the requirements is saved so that the application can be deployed in the actual system to replace the original model. This embodiment can effectively improve the accuracy of the big model's intention recognition on claims issues, thereby optimizing the pertinence and effectiveness of responses, improving the overall efficiency and quality of claims services, and enhancing user experience. At the same time, it also allows the big model to better adapt to the situation in the claims field, helping companies improve their competitiveness and save related resource costs.

[0075] In some embodiments, if the first judgment result is no, the method further includes: obtaining a reply intention and a related knowledge table link for the user intention input; and expanding the script of the dialogue script set according to the user intention, the reply intention, and the related knowledge table link.

[0076] For example, a user sends a claim problem to be resolved, "My medical insurance claim application has been submitted for a long time, why hasn't it been reviewed and approved yet?" The identified user intent is "to inquire about the reason why the medical insurance claim application was not reviewed and approved." If there is no best matching intent in the dialogue script, the business personnel of the claims company can enter the reply intent of "explain to the user the review time of the medical insurance claim application and the common reasons why it may not be passed" and the claim process table link through the front-end interface. The user intent, reply intent, and claim process table link in this example are added to the corresponding columns in the dialogue script set to expand the dialogue script set. This embodiment can expand the dialogue script set based on the user's actual intentions and corresponding replies, knowledge links and other information to make the dialogue script more comprehensive and detailed, thereby improving the accuracy and professionalism of the replies, optimizing the interactive experience between users and customer service, and enhancing the ability of the entire claims customer service system to respond to diverse user needs.

[0077] In some embodiments, the method further includes: if the matching degree corresponding to the best matching question is less than a preset matching threshold, obtaining the question to be supplemented and the answer to be supplemented corresponding to the user intention; and expanding the question and answer form according to the question to be supplemented and the answer to be supplemented.

[0078] Here, the preset matching threshold can be set according to actual needs, for example, set to 0.8. If the matching degree corresponding to the best matching question is less than 0.8, it means that the questions in the question and answer table are too far from the semantics of the user's intention. At this time, the questions and corresponding reply texts input by the business personnel of the claims company through the front-end interface for the user's intention can be received, and the questions and reply texts can be added to the question and answer table. This embodiment can supplement the content of the question and answer table in a timely manner, so that the question and answer table is continuously enriched and improved, and can then cover more types of specific questions that more users may raise, improve the comprehensiveness and accuracy of the system's responses, better respond to diverse user inquiries, and enhance the ability of the entire claims consulting service to respond to complex situations. It effectively improves the user experience and allows users to obtain the required claims-related information more conveniently and accurately.

[0079] The present disclosure also provides an intelligent question-answering processing method, comprising the following steps:

[0080] Step S1: Obtain the unresolved claims issue sent by the user, call the large model interface, perform intent recognition on the unresolved claims issue, and obtain the user's intent.

[0081] Step S2: Obtain information on the type of insurance involved in the claim settlement issue to be resolved.

[0082] Step S3: Filter the scripts in the dialogue script set according to the insurance type information to obtain a target dialogue script set.

[0083] Among them, the method for generating a set of dialogue scripts is as follows: obtain multiple groups of historical dialogues between users and claims adjusters; for each group of target dialogues in the multiple groups of historical dialogues between users and claims adjusters, use a large model to perform intent recognition on the dialogue text in the target dialogue, and obtain a target dialogue script corresponding to the target dialogue; perform duplication elimination processing on any unprocessed dialogue scripts in each target dialogue script to obtain a set of dialogue scripts; wherein, performing duplication elimination processing on the dialogue scripts to be processed includes: for any target user intention in the dialogue script to be processed, respectively calculate the similarity between the target user intention and each user intention in other target dialogue scripts to obtain a target similarity; if the target similarity is greater than a preset similarity, delete the target user intention and the corresponding claims adjuster's reply intention, or delete the corresponding user intention and claims adjuster's reply intention in other target dialogue scripts.

[0084] Step S4: Calculate the matching degree between the user intention and the intention of each user role in each script in the target dialogue script set respectively to obtain multiple matching degrees.

[0085] Step S5: Based on the preset screening conditions and multiple matching degrees, determine whether there is a best matching intention and obtain a first judgment result.

[0086] Step S6: If the first judgment result is yes, then by calling the large model interface, according to the best matching intention, the corresponding associated knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and the process returns to the step of obtaining the claims problem to be resolved sent by the user.

[0087] Step S7: If the first judgment result is no, then the matching degree between the user intention and each question in the question-and-answer table is calculated.

[0088] If the first judgment result is no, the reply intention and related knowledge table link for the user intention input are also obtained; according to the user intention, the reply intention and the related knowledge table link, the script of the dialogue script set is expanded.

[0089] Step S8: for the largest preset number of matching degrees to be screened among the matching degrees, using the preset difference judgment condition, it is judged whether the differences between the preset number of matching degrees to be screened are large, and a second judgment result is obtained.

[0090] Step S9: If the second judgment result is no, the large model is used to select the best matching question from a preset number of questions corresponding to the matching degree to be screened.

[0091] Step S10: If the second judgment result is yes, the question corresponding to the maximum matching degree to be screened is taken as the best matching question.

[0092] Step S11: The answer to the best matching question is used as the reply text.

[0093] If the matching degree corresponding to the best matching question is less than a preset matching threshold, the question to be supplemented and the answer to be supplemented corresponding to the user's intention are obtained; and the question and answer table is expanded according to the question to be supplemented and the answer to be supplemented.

[0094] Those skilled in the art will appreciate that, in the above method of the specific implementation manner, the writing order of the steps does not mean a strict execution order, but does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0095] It should be noted that, in practical applications, all possible implementations described above can be combined in any manner to form possible embodiments of the present disclosure, which will not be described one by one here.

[0096] It should be noted that the information (including but not limited to device information, user information, etc.) and data (including but not limited to data used for analysis, storage and display, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. In addition, the numbers corresponding to the various steps in the above embodiments only serve as identification and are not used as a limitation on the order in which the steps are executed. The order in which the steps are executed in each embodiment can be set according to actual conditions.

[0097] Based on the same concept, the embodiment of the present disclosure also provides an intelligent question and answer processing device, which corresponds one-to-one to the intelligent question and answer processing method in the above embodiment. Figure 3 FIG. 1 shows a schematic diagram of the structure of the intelligent question-answering processing device provided in an embodiment of the present disclosure, see Figure 3 As shown, the intelligent question-answering processing device 300 provided in the embodiment of the present disclosure includes:

[0098] The intention recognition module 301 is used to obtain the pending claims sent by the user, and call the large model interface to perform intention recognition on the pending claims to obtain the user intention;

[0099] The reply module 302 is used to match the user intention with the intention of each user role in the script in the dialogue script set, respectively, to obtain multiple matching degrees; and based on the preset screening conditions and the multiple matching degrees, determine whether there is a best matching intention to obtain a first judgment result; wherein the content of the script in the dialogue script set includes dialogue roles, role intentions and related knowledge table links; if the first judgment result is yes, then by calling the big model interface, according to the best matching intention, the corresponding related knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and returns to the step of obtaining the unresolved claims problem sent by the user; if the first judgment result is no, then the big model interface is called, and a reply text is generated and returned to the user according to the question and answer table or the knowledge table, and the user intention.

[0100] It should be noted that the device is implemented based on a claims knowledge base, which includes a question and answer table, a knowledge table and a dialogue script set. The knowledge table includes a policy information table, a claims process table, a product knowledge table, and a regulations and policy table.

[0101] In some embodiments, in the above-mentioned device, the reply module 302 is used to: obtain the insurance type information involved in the claim issue to be resolved; screen the scripts in the dialogue script set according to the insurance type information to obtain a target dialogue script set; and respectively calculate the matching degree between the user intention and the intention of each user character in each script in the target dialogue script set to obtain the multiple matching degrees.

[0102] In some embodiments, the device also includes a script generation module, which is used to: obtain multiple groups of historical conversations between users and claims adjusters; for each group of target conversations in the multiple groups of historical conversations between users and claims adjusters, use a large model to perform intent recognition on the conversation text in the target conversation to obtain a target conversation script corresponding to the target conversation; perform duplication elimination processing on any pending conversation script in each of the target conversation scripts to obtain the conversation script set; wherein, performing duplication elimination processing on the pending conversation scripts includes: for any target user intention in the pending conversation script, respectively calculating the similarity between the target user intention and each user intention in other target conversation scripts to obtain a target similarity; if the target similarity is greater than a preset similarity, deleting the target user intention and the corresponding claims adjuster response intention, or deleting the corresponding user intention and claims adjuster response intention in other target conversation scripts.

[0103] In some embodiments, in the above-mentioned device, the reply module 302 is used to: calculate the matching degree between the user intention and each question in the question-and-answer table; for the largest preset number of matching degrees to be screened among each of the matching degrees, use a preset difference judgment condition to judge whether the difference between the preset number of matching degrees to be screened is large, and obtain a second judgment result; if the second judgment result is no, use a large model to select the best matching question from the questions corresponding to the preset number of matching degrees to be screened; if the second judgment result is yes, use the question corresponding to the largest matching degree to be screened as the best matching question; and use the answer to the best matching question as the reply text.

[0104] In some embodiments, the device also includes a model optimization module, which is used to: obtain multiple unresolved claims issues where intent recognition fails; perform correct intent labeling on the multiple unresolved claims issues where intent recognition fails, and obtain a fine-tuning sample set; use the fine-tuning sample set to fine-tune the large model, and replace the large model with the fine-tuned large model.

[0105] In some embodiments, if the first judgment result is no, the device also includes a script optimization module, which is used to: obtain a reply intention and a related knowledge table link for the user intention input; and expand the script of the dialogue script set according to the user intention, the reply intention, and the related knowledge table link.

[0106] In some embodiments, the device also includes a question and answer knowledge expansion module, which is used to: if the matching degree corresponding to the best matching question is less than a preset matching threshold, obtain the questions to be supplemented and the answers to be supplemented corresponding to the user intention; and expand the question and answer table according to the questions to be supplemented and the answers to be supplemented.

[0107] The present invention provides an intelligent question-and-answer processing device, which operates based on a claims knowledge base including a question-and-answer table, a knowledge table (including a policy information table, a claims process table, a product knowledge table, and a regulations and policy table) and a dialogue script set. First, the unresolved claims problem sent by the user is obtained, and the large model interface is called for intent recognition. The user's intention is deeply understood by relying on the powerful natural language processing ability of the large model, and the subtle differences are accurately captured, laying the foundation for subsequent accurate matching of user needs. Then, with the help of the dialogue script, a clear and organized communication framework is built for the user through the predefined dialogue process, and the user is guided to have a structured dialogue, ensuring that the service is consistent from the beginning, and can be flexibly adjusted according to the user's specific questions and the dialogue context in which it is located, providing guidance that fits the user's personalized needs, which not only effectively reduces the communication confusion caused by the lack of guidance of the user, but also reduces the dependence on manual customer service and the corresponding training costs. When the dialogue script cannot handle the user's problem, the large model is used to retrieve relevant information from the question-and-answer table or knowledge table, and a reply text is generated based on the relevant information. The accuracy and stability of the question-and-answer table or knowledge table are crucial to reducing human errors and improving service reliability. By closely combining and coordinating the three elements of big model intention understanding, dialogue script guidance, and claims knowledge base, this solution has significant advantages over traditional online customer service solutions. It can reduce labor costs, improve response speed, ensure information reliability, enhance user experience, and guide smooth communication. It performs well in multiple dimensions and can comprehensively improve customer service quality and efficiency, helping companies save resource costs and enhance competitiveness. It can also improve customer retention rate and sales conversion rate through personalized recommendations and care, create greater business value, meet user claims needs, and consolidate the company's service advantage.

[0108] For the specific definition of the intelligent question and answer processing device, please refer to the definition of the intelligent question and answer processing method above, which will not be repeated here. Each module in the above-mentioned intelligent question and answer processing device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0109] Figure 4 FIG. 1 shows a schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure. Figure 4As shown, at the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. Among them, the memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage, etc. Of course, the electronic device may also include hardware required for other services.

[0110] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0111] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0112] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming an intelligent question-answering processing device at the logical level. The processor executes the program stored in the memory and is specifically used to execute the above method.

[0113] The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present disclosure may be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present disclosure may be directly embodied as being executed by a hardware decoding processor, or may be executed by a combination of hardware and software modules in a decoding processor. The software module may be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0114] The electronic device can execute the intelligent question-answering processing method provided by the multiple embodiments of the present disclosure, and realize the intelligent question-answering processing device in Figure 3 The functions of the illustrated embodiments will not be described in detail in the embodiments of the present disclosure.

[0115] The embodiments of the present disclosure also propose a computer-readable storage medium, which stores one or more programs, and the one or more programs include instructions. When the instructions are executed by an electronic device including multiple application programs, the electronic device can execute the intelligent question and answer processing method provided by multiple embodiments of the present disclosure.

[0116] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0120] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0121] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0122] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0123] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity or device including the element.

[0124] It will be appreciated by those skilled in the art that the embodiments of the present disclosure may be provided as methods, systems or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] The above are only embodiments of the present disclosure and are not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the scope of the claims of the present disclosure.

Claims

1. An intelligent question-answering processing method, characterized in that: The method is implemented based on a claims knowledge base, which includes a question-and-answer table, a knowledge table, and a dialogue script set. The knowledge table includes a policy information table, a claims process table, a product knowledge table, and a regulations and policies table. The method includes: Obtain the pending claims sent by the user, and call the big model interface to perform intent recognition on the pending claims to obtain the user's intent; The user intention is matched with the intention of each user role in the script of the dialogue script collection to obtain multiple matching degrees; and according to the preset screening conditions and the multiple matching degrees, it is judged whether there is a best matching intention to obtain a first judgment result; wherein the content of the script in the dialogue script collection includes dialogue roles, role intentions and related knowledge table links; If the first judgment result is yes, then by calling the big model interface, according to the best matching intention, the corresponding associated knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and the process returns to the step of obtaining the claims problem to be solved sent by the user; If the first judgment result is no, the large model interface is called to generate a reply text based on the question and answer table or the knowledge table and the user's intention and return it to the user.

2. The method according to claim 1, characterized in that The user intention is matched with the intention of each user role in the script of the dialogue script set to obtain multiple matching degrees, including: Obtaining information on the type of insurance involved in the claim settlement issue to be resolved; According to the insurance type information, the scripts in the dialogue script set are screened to obtain a target dialogue script set; The matching degree between the user intention and the intention of each user role in each script in the target dialogue script set is calculated respectively to obtain the multiple matching degrees.

3. The method according to claim 1, characterized in that The dialogue script set is generated according to the following method: Get multiple groups of historical conversations between users and claims adjusters; For each target dialogue in the multiple groups of historical dialogues between users and claims adjusters, use the big model to perform intent recognition on the dialogue text in the target dialogue to obtain a target dialogue script corresponding to the target dialogue; Performing repetitive elimination processing on any dialogue script to be processed in each of the target dialogue scripts to obtain the dialogue script set; The process of removing duplicate content from the dialogue script to be processed includes: For any target user intention in the dialogue script to be processed, respectively calculate the similarity between the target user intention and each user intention in other target dialogue scripts to obtain a target similarity; If the target similarity is greater than a preset similarity, the target user intent and the corresponding claims adjuster's response intent are deleted, or the corresponding user intent and claims adjuster's response intent in other target dialogue scripts are deleted.

4. The method according to claim 1, characterized in that: Calling the big model interface to generate a reply text according to the question-answer form and the user's intention, including: Calculate the matching degree between the user intention and each question in the question-and-answer table; For the largest preset number of matching degrees to be screened among the matching degrees, using a preset difference judgment condition, it is judged whether the differences between the preset number of matching degrees to be screened are large, and a second judgment result is obtained; If the second judgment result is no, then using the large model to select the best matching question from the preset number of questions corresponding to the matching degree to be screened; If the second judgment result is yes, the question corresponding to the maximum matching degree to be screened is taken as the best matching question; The answer to the best matching question is used as the reply text.

5. The method according to claim 1, characterized in that The method further comprises: Get multiple pending claims that failed intent recognition; For the multiple claims issues that failed to be resolved, respectively, correct intent labeling is performed to obtain a fine-tuning sample set; The large model is fine-tuned using the fine-tuning sample set, and the large model is replaced by the fine-tuned large model.

6. The method according to claim 1, characterized in that If the first judgment result is no, the method further includes: Obtaining a reply intent and a related knowledge table link for the user intent input; The script of the dialogue script set is expanded according to the user intention, the reply intention and the related knowledge table link.

7. The method according to any one of claims 4 to 6, characterized in that: The method further comprises: If the matching degree corresponding to the best matching question is less than a preset matching threshold, obtaining the question to be supplemented and the answer to be supplemented corresponding to the user's intention; The question and answer table is expanded according to the questions to be supplemented and the answers to be supplemented.

8. An intelligent question-answering processing device, characterized in that: The device is implemented based on a claims knowledge base, which includes a question-and-answer table, a knowledge table, and a dialogue script set. The knowledge table includes a policy information table, a claims process table, a product knowledge table, and a regulations and policies table. The device comprises: The intention recognition module is used to obtain the pending claims sent by the user, and call the large model interface to perform intention recognition on the pending claims to obtain the user's intention; A reply module is used to match the user intention with the intention of each user role in the script in the dialogue script set, respectively, to obtain multiple matching degrees; and based on the preset screening conditions and the multiple matching degrees, determine whether there is a best matching intention to obtain a first judgment result; wherein the content of the script in the dialogue script set includes dialogue roles, role intentions and related knowledge table links; if the first judgment result is yes, then by calling the big model interface, according to the best matching intention, the corresponding related knowledge table link and the claims adjuster role, a guiding reply text or an answer reply text is generated and returned to the user, and the step of obtaining the unresolved claims problem sent by the user is returned; if the first judgment result is no, then the big model interface is called, and a reply text is generated and returned to the user according to the question and answer table or the knowledge table, and the user intention.

9. An electronic device, comprising: processor; as well as A memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to perform the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing one or more programs, characterized in that: When the one or more programs are executed by an electronic device including a plurality of application programs, the electronic device executes the steps of the method as claimed in any one of claims 1 to 7.

Images