Intelligent customer service response method and system based on large model

Abstract

The present invention discloses an intelligent customer service answering method and system based on a large model, wherein the method comprises: obtaining user input and the user's historical dialogue, and splicing them with a preset first prompt, inputting the large model, so as to output query rewriting and intent recognition; responding to intent recognition, and splicing the intent recognition with a preset second prompt, inputting the large model, so as to output a call instruction; responding to the call instruction, and generating content output. By obtaining the content input by the user and combining it with the user's corresponding historical dialogue, the processing content for the user can be automatically completed, and the prompt generated by the large model through the processing content can effectively rewrite and recognize the intent of the processing content. In this way, the requirements for the user input content are greatly reduced, and the processing process is more flexible. At the same time, after obtaining accurate intent recognition, the next step of processing can be performed more accurately to obtain accurate content output, and the efficiency and accuracy of processing user problems are higher.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent customer service, and in particular to an intelligent customer service response method and system based on a large model. Background Art

[0002] With the acceleration of the Internet of Everything, users can consult the information they need or other matters through the Internet. Traditional ports for user interaction usually use static data display, and users need to perform a large number of programmed operations according to their needs. Therefore, some question-and-answer type intelligent customer service has emerged, but the current intelligent customer service has great restrictions on the user input demand paradigm, and the user's use threshold is very high. Therefore, intelligent customer service robots that can recognize the user's input content intention have great development value. Summary of the invention

[0003] In view of this, the present invention proposes an intelligent customer service response method and system based on a large model, which can accurately identify the user's intention and output a highly accurate answer. The present invention provides the following technical solutions:

[0004] An intelligent customer service response method based on a large model, the method comprising:

[0005] Get user input and historical conversations, concatenate them with the preset first prompt, and input them into the big model to output query rewriting and intent recognition;

[0006] Respond to the intention recognition, and combine the intention recognition with the preset second prompt, input it into the big model, and output the call instruction;

[0007] Respond to invocation instructions and generate content output.

[0008] Furthermore, the method of obtaining user input and the user's historical dialogue, splicing them with the preset first prompt, inputting into the large model, and outputting query rewriting and intent recognition includes:

[0009] Get user tags based on user input;

[0010] Query historical conversations based on user tags;

[0011] If there is a historical conversation, the historical conversation, user input and the first prompt are concatenated and input into the big model. The big model rewrites the query based on the historical conversation and user input, and outputs the rewritten query and intent recognition result.

[0012] Otherwise, the user input and the first prompt are directly concatenated and input into the large model to output the rewritten query and intent recognition.

[0013] Furthermore, the intention recognition includes consultation, business inquiry and processing, transfer to manual service and chatting.

[0014] Further, when the intention recognition is consultation, the response intention recognition is performed, and the intention recognition is spliced ​​with the preset second prompt, and input into the large model to output the call instruction; the method of responding to the call instruction and generating content output includes:

[0015] In response to the inquiry intent, the inquiry intent and the rewritten query are concatenated with the preset second prompt, and input into the big model to output the instruction to call the embedding model;

[0016] In response to the instruction to call the embedding model, the rewritten query is converted into a question vector through the embedding model and searched in the knowledge base;

[0017] If no search results are generated, the fixed script will be output and transferred to manual work;

[0018] Otherwise, the search results are recalled and the preset knowledge base question and answer instructions are spliced, and input into the big model to output the consultation answers and generate content output.

[0019] Further, when the intention is recognized as business inquiry and processing, the response intention is recognized, and the intention recognition is spliced ​​with the preset second prompt, input into the large model, so as to output the call instruction; the method of responding to the call instruction and generating content output includes:

[0020] Respond to business inquiries and processing intentions, and determine the classification of current business inquiries and processing intentions based on the rewritten query;

[0021] Confirm the fields to be collected based on user input, user history, and current classification;

[0022] Complete field information extraction according to the fields that need to be collected, splice the extracted field information with the preset second prompt, and input it into the big model to output data query instructions;

[0023] Obtain the query results of the field data to be collected through data query instructions;

[0024] The query results are spliced ​​with the preset third prompt and input into the big model to rewrite the query results and generate content output.

[0025] Furthermore, the method of obtaining the query result of the field data to be collected through the data query instruction includes:

[0026] Update the extracted field information to the collection information field library;

[0027] Whenever the collected information field library is updated, determine whether there are empty fields and confirm the number of questions corresponding to each empty field;

[0028] When the number of questions about an empty field is less than two, select the corresponding empty fields in turn, concatenate the empty fields with the preset fourth prompt, and input the large model to output an empty field question instruction, ask questions to the user, and update the corresponding empty fields after receiving the reply;

[0029] When there is no empty field or the number of questions about an empty field is greater than or equal to two, the confirmed field information is summarized;

[0030] Perform data query on offline database based on the summarized field information and obtain query results.

[0031] Furthermore, the method of performing data query on an offline database according to the aggregated field information and obtaining the query result includes:

[0032] Determine whether the user is an account manager based on the current user's tag.

[0033] If yes, then perform data query in the offline database according to the summarized field information and obtain the query result;

[0034] Otherwise, the information of the current user is reviewed. If the user is confirmed to be the person with the current user tag, data query is performed on the offline database based on the summarized field information and the query results are obtained. Otherwise, it is transferred to manual.

[0035] The present invention further discloses an intelligent customer service response system based on a large model, comprising:

[0036] The input module obtains user input and historical user conversations, concatenates them with the preset first prompt, and inputs them into the large model to output query rewriting and intent recognition;

[0037] The processing module responds to the intention recognition, and splices the intention recognition with the preset second prompt, inputs it into the large model, and outputs the call instruction;

[0038] The output module responds to the call instructions and generates content output.

[0039] The present invention further discloses a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0040] The present invention further discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the program.

[0041] According to the technical solution of the present invention, by obtaining the content input by the user and combining it with the user's corresponding historical conversations, the processing content for the user can be automatically completed. The large model can effectively rewrite and identify the intent of the processing content through the processing content combined with the prompt. This method greatly reduces the requirements for user input content, and the processing process is more flexible. At the same time, after obtaining accurate intent recognition, the next step of processing can be performed more accurately to obtain accurate content output, and the efficiency and accuracy of processing user problems are higher. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] For the purpose of illustration and not limitation, the present invention is now described in conjunction with the embodiments of the present invention and the accompanying drawings, in which:

[0043] Figure 1 It is a flowchart of a large model-based intelligent customer service response method in an embodiment of the present invention;

[0044] Figure 2 It is a schematic diagram of the processing flow of the intelligent customer service response system based on a large model in an embodiment of the present invention;

[0045] Figure 3 It is a structural diagram of an intelligent customer service response system based on a large model in an embodiment of the present invention;

[0046] Figure 4 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the implementation mode of the present application will be clearly and completely described below in conjunction with the drawings in the implementation mode of the present application. Obviously, the described implementation mode is only a part of the implementation mode of the present application, not all the implementation modes. Based on the implementation mode in the present application, all other implementation modes obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application.

[0048] refer to Figure 1 and Figure 2 This embodiment discloses a method for intelligent customer service response based on a large model, which is applied to various business scenarios. The business scenarios are configured with large models, such as language models and vector models. Among them, the language model has the capabilities of language understanding, logical reasoning, content generation, knowledge question answering, etc. The vector model converts text strings into vectors, so that knowledge data can be effectively used in search, clustering, recommendation systems, anomaly detection and classification, etc. This embodiment uses the power business as an example to disclose this technical solution. The specific method includes:

[0049] S100: Obtain user input and the user's historical conversations, and concatenate them with the preset first prompt, input them into the big model, and output query rewriting and intent recognition.

[0050] This step is based on the language model in the business scenario. The user can input any topic without the restriction of input paradigm. The user input is more free and the threshold for using related services is lower. The historical conversation is the historical usage record of the current user, which is stored in the database and automatically extracted when the user triggers input to realize the identity recognition of the user. The details are as follows:

[0051] S110: Obtaining a user tag according to user input;

[0052] S120: Query historical conversations based on user tags;

[0053] S130: If there is a historical conversation, the historical conversation, the user input and the first prompt are concatenated and input into the large model. The large model rewrites the query according to the historical conversation and the user input, and outputs the rewritten query and the intent recognition result.

[0054] S140: If there is no historical dialogue, the user input and the first prompt are directly concatenated and input into the large model to output the rewritten query and intent recognition.

[0055] Among them, the user tag is a tag that specifically refers to a single user. Exemplarily, user A enters a balance query, and the tag of user A - mobile phone number is retrieved. Based on the mobile phone number, the account number obtained by user A when using it before is extracted from the database. The user's tag is not fixed, and it also includes other tags that can specifically refer to the user, such as email addresses. Then the information obtained by this large model-based intelligent customer service is: User A: Account number is xxx, User A: Check balance, based on this splicing, the preset first prompt is spliced, and the query is rewritten: Account number: xxx, check balance. Similarly, when there is no usage record of the relevant user stored in the database, it will not be retrieved, and the query rewrite will be directly output through the first prompt. The historical conversation of the user who has no usage record can also be set to an empty field. This large model-based intelligent customer service performs intent recognition based on the first prompt.

[0056] In this embodiment, intent recognition is divided into four categories, including, for example, consultation, business inquiry and processing, transfer to manual service, and small talk.

[0057] S200: Respond to the intention recognition, and combine the intention recognition with the preset second prompt, input into the large model, and output the call instruction.

[0058] Specific:

[0059] When the intent is identified as a consultation, in response to the consultation intent, the consultation intent and the rewritten query are concatenated with the preset second prompt and input into the large model to output the instruction for calling the embedding model.

[0060] When the intent is identified as business inquiry and processing, respond to the business inquiry and processing intent, and determine the classification of the current business inquiry and processing intent based on the rewritten query;

[0061] Confirm the fields to be collected based on user input, user history, and current classification;

[0062] Complete field information extraction based on the fields that need to be collected, splice the extracted field information with the preset second prompt, and input it into the big model to output data query instructions.

[0063] When the intention is recognized as manual transfer, respond to the intention to transfer to manual, identify the specific classification labels in the intention to transfer to manual, such as complaints, opinions, repairs, praise, reports, service applications, suggestions, reminders, etc., query the above labels and identify whether the labels are in processing. If so, concatenate the classification label with the preset second prompt and input the large model to retrieve the historical data corresponding to the label, package the historical data and the customer input data for manual transfer. If not, concatenate the classification label with the preset second prompt and input the large model to output the words corresponding to the current label, and finally output the content reply.

[0064] When the intent is identified as small talk, the backup script is directly retrieved and output.

[0065] It should be noted that the above-mentioned multiple second prompts are not the same, that is, when the types of intended recognition are different, the contents of the spliced ​​preset second prompts are also different.

[0066] S300: Respond to the call instruction and generate content output.

[0067] That is, when the intent is identified as consultation, the response is to call the embedding model instruction;

[0068] S311: responding to the instruction to call the embedding model, converting the rewritten query into a question vector through the embedding model, and searching in the knowledge base;

[0069] S312: If no search results are generated, output the fixed script and transfer to manual work;

[0070] S313: Otherwise, recall the search results and splice the preset knowledge base question and answer instructions, input the big model to output the consultation answer and generate content output.

[0071] When the intention is identified as business inquiry and processing, respond to the data query instruction;

[0072] S320: Obtaining query results of the field data to be collected through the data query instruction;

[0073] S330: The query result is spliced ​​with the preset third prompt and input into the big model to rewrite the query result and generate content output.

[0074] Wherein, step S320 specifically includes:

[0075] S321: updating the extracted field information to the collection information field library;

[0076] S322: Whenever the collected information field library is updated, determine whether there are empty fields and confirm the number of questions corresponding to each empty field;

[0077] S323: When the number of questions about the empty field is less than two, select the corresponding empty fields in turn, splice the empty fields with the preset fourth prompt, and input the large model to output the empty field questioning instruction, initiate the question to the user, and update the corresponding empty field after receiving the reply;

[0078] S324: When there is no empty field or the number of questions about the empty field is greater than or equal to two, summarize the confirmed field information;

[0079] S325: Perform data query on the offline database according to the summarized field information and obtain the query result.

[0080] Further, step S325 specifically includes:

[0081] Determine whether the user is an account manager based on the current user's tag.

[0082] If yes, then perform data query in the offline database according to the summarized field information and obtain the query result;

[0083] Otherwise, the information of the current user is reviewed. If the user is confirmed to be the person with the current user tag, data query is performed on the offline database based on the summarized field information and the query results are obtained. Otherwise, it is transferred to manual.

[0084] Among them, when judging whether the user is an account manager, this embodiment uses the current user's mobile phone number as the basis for judgment. If the mobile phone number is judged, the data query is performed in the offline database according to the summarized field information, and the query result is obtained. Otherwise, the current user's information is reviewed, for example, an identity confirmation card pops up. If the customer clicks yes, that is, the current user is an account manager, the data query is performed in the offline database according to the summarized field information, and the query result is obtained, otherwise it is transferred to manual.

[0085] After the question and answer is completed, the method of this embodiment also includes: completing the knowledge base slice recall according to the user input, collecting the user query and the corresponding slice, retrieving the splicing knowledge base question and answer instruction, and conducting the knowledge base question and answer.

[0086] It should be understood that when a user interacts with an intelligent customer service, there will be at least one round of conversation. When there are multiple rounds of conversations, the recognition of the user's intent will change each time. Therefore, the response methods corresponding to different intent recognitions in this embodiment can be combined with each other in a user's complete interaction process.

[0087] This embodiment further discloses a system architecture for implementing the above-mentioned intelligent customer response method, specifically: including access layer, application layer, platform layer, operating environment and business system. Among them, access layer: including management background, APP and OpenAp i, serving internal employees, C-end customers and interface calls between systems respectively. Application layer: composed of intelligent customer service subsystem, analysis report subsystem and digital human subsystem, realizing functions such as intelligent question and answer, business query, and multi-round dialogue. Platform layer: based on the basic information platform and the large model platform, it provides general middle-end capabilities and general capabilities of large models. Operating environment: adopts microservice architecture, including server, nginx, relational database, vector database, cache database to meet the requirements of information innovation. Simulation business system: simulates customer data, such as customer information, electricity bill and power information, power outage plan, work order information, etc.

[0088] refer to Figure 3 This embodiment further discloses an intelligent customer service response system based on a large model, including:

[0089] The input module 21 is used to obtain user input and the user's historical conversations, and concatenate them with the preset first prompt, and input them into the large model to output query rewriting and intent recognition. Specifically, the user tag is obtained according to the user input; the historical conversation is queried according to the user tag; if there is a historical conversation, the historical conversation, the user input and the first prompt are concatenated and input into the large model, and the large model rewrites the query according to the historical conversation and the user input, and outputs the rewritten query and intent recognition result; otherwise, the user input and the first prompt are directly concatenated and input into the large model to output the rewritten query and intent recognition.

[0090] Processing module 22 is used to respond to intent recognition, and splice the intent recognition with the preset second prompt, input the large model, and output the call instruction; specifically: when the intent is recognized as consultation, respond to the consultation intent, splice the consultation intent and the rewritten query with the preset second prompt, input the large model, and output the call embedding model instruction. When the intent is recognized as business inquiry and processing, respond to the business inquiry and processing intent, and determine the classification of the current business inquiry and processing intent according to the rewritten query; confirm the fields that need to be collected according to the user input, the user's historical conversations and the current classification; complete the field information extraction according to the fields that need to be collected, splice the extracted field information with the preset second prompt, and input the large model to output the data query instruction. When the intent is identified as manual transfer, respond to the intent to transfer to manual, identify the specific classification tags in the intent to transfer to manual, such as complaints, opinions, repairs, praise, reports, service applications, suggestions, and urging, query the above tags and identify whether the tags are in process. If so, concatenate the classification tag with the preset second prompt and input the large model to retrieve the historical data corresponding to the tag, package the historical data and customer input data for manual transfer. If not, concatenate the classification tag with the preset second prompt and input the large model to output the speech corresponding to the current tag, and finally output the content reply. When the intent is identified as small talk, directly retrieve the backup speech and output it.

[0091] The output module 23 is used to respond to the call instruction and generate content output; specifically, when the intention is recognized as consultation, it responds to the call embedding model instruction; responds to the call embedding model instruction, converts the rewritten query into a question vector through the embedding model, and searches in the knowledge base; if no search results are generated, the fixed words are output and transferred to manual; otherwise, the search results are recalled and spliced ​​with the preset knowledge base question and answer instructions, and input into the big model to output the consultation answer and generate content output. When the intention is recognized as business inquiry and processing, it responds to the data query instruction; obtains the query results of the field data to be collected through the data query instruction; splices the query results with the preset third prompt, and inputs into the big model to rewrite the query results and generate content output. Update the extracted field information to the collection information field library; whenever the collection information field library is updated, determine whether there are empty fields and confirm the number of questions corresponding to each empty field; when the number of questions about an empty field is less than two, select the corresponding empty fields in turn, splice the empty fields with the preset fourth prompt, and input the large model to output the empty field question instruction, ask the user a question, and update the corresponding empty field after receiving the reply; when there is no empty field or the number of questions about an empty field is greater than or equal to two, summarize the confirmed field information; perform data query in the offline database based on the summarized field information and obtain the query results. Determine whether the user is an account manager based on the current user's label. If so, perform data query in the offline database based on the summarized field information and obtain the query results; otherwise, review the information of the current user. If the user is confirmed to be the person with the current user label, perform data query in the offline database based on the summarized field information and obtain the query results, otherwise, transfer to manual.

[0092] In summary, by obtaining the content input by the user and combining it with the user's corresponding historical dialogue, the processing content for the user can be automatically completed. The large model can effectively rewrite and identify the intent of the processing content through the processing content and the preset prompt. This method greatly reduces the requirements for user input content, and the processing process is more flexible. At the same time, after obtaining accurate intent recognition, the next step of processing can be performed more accurately to obtain accurate content output, and the efficiency and accuracy of processing user problems are higher. In experimental verification, the intelligent customer service model call success rate in this implementation is >98%, and the average time taken for each round of dialogue response is <3.5s, that is: the first ring time of each round of dialogue of the large model is <3.5s; the tp95 time taken for each round of dialogue is <10s.

[0093] refer to Figure 4, is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, the electronic device 50 includes: a processor 501 (processor), a memory 502 (memory) and a bus 503;

[0094] The processor 501 and the memory 502 communicate with each other via the bus 503 ; the processor 501 is used to call program instructions in the memory 502 to execute the methods provided by the above-mentioned method implementation methods.

[0095] This embodiment provides a non-transitory computer-readable storage medium, which stores computer instructions. The computer instructions enable a computer to execute the methods provided by the above-mentioned method embodiments.

[0096] It should be noted that the terms "first", "second", etc. in the specification of the present application 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 the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0097] In the present application, the terms "upper", "lower", "left", "right", "front", "back", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings. These terms are mainly used to better describe the present application and its implementation methods, and are not used to limit the indicated devices, elements or components to have a specific orientation, or to be constructed and operated in a specific orientation.

[0098] In addition, some of the above terms may be used to express other meanings in addition to indicating orientation or positional relationship. For example, the term "on" may also be used to express a certain dependency or connection relationship in some cases. For those of ordinary skill in the art, the specific meanings of these terms in this application can be understood according to the specific circumstances. In addition, the term "plurality" should mean two or more than two.

[0099] It should be noted that, in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0100] A person skilled in the art can understand that all or part of the steps for implementing the above-mentioned method implementation method can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium, which, when executed, executes the steps of the above-mentioned method implementation method; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various storage media that can store program codes.

[0101] The device implementation described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present implementation scheme. Those of ordinary skill in the art may understand and implement it without creative effort.

[0102] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each implementation method or some parts of the implementation method.

[0103] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent customer service response method based on a large model, characterized in that: The method comprises: Get user input and historical conversations, concatenate them with the preset first prompt, and input them into the big model to output query rewriting and intent recognition; Respond to the intention recognition, and combine the intention recognition with the preset second prompt, input it into the big model, and output the call instruction; Respond to invocation instructions and generate content output.

2. The intelligent customer service response method based on a large model according to claim 1 is characterized in that: The method of obtaining user input and the user's historical dialogue, splicing them with the preset first prompt, and inputting them into the large model to output query rewriting and intent recognition includes: Get user tags based on user input; Query historical conversations based on user tags; If there is a historical conversation, the historical conversation, user input and the first prompt are concatenated and input into the big model. The big model rewrites the query based on the historical conversation and user input, and outputs the rewritten query and intent recognition result. Otherwise, the user input and the first prompt are directly concatenated and input into the large model to output the rewritten query and intent recognition.

3. The intelligent customer service response method based on a large model according to claim 1 is characterized in that: The intention recognition includes consultation, business inquiry and processing, transfer to manual service and chatting.

4. The intelligent customer service response method based on a large model according to claim 3 is characterized in that: When the intention is recognized as consultation, the response intention is recognized, and the intention recognition is spliced ​​with the preset second prompt, and input into the large model to output the call instruction; The method for responding to the call instruction and generating content output includes: In response to the inquiry intent, the inquiry intent and the rewritten query are concatenated with the preset second prompt and input into the big model to output the instruction to call the embedding model. In response to the embedding model instruction, the rewritten query is converted into a question vector through the embedding model and retrieved in the knowledge base; If no search results are generated, the fixed script will be output and transferred to manual work; Otherwise, the search results are recalled and the preset knowledge base question and answer instructions are spliced, and input into the big model to output the consultation answers and generate content output.

5. The intelligent customer service response method based on a large model according to claim 3 is characterized in that: When the intention is recognized as business inquiry and processing, the response intention is recognized, and the intention recognition is spliced ​​with the preset second prompt and input into the big model to output the call instruction; The method for responding to the call instruction and generating content output includes: Respond to business inquiries and processing intentions, and determine the classification of current business inquiries and processing intentions based on the rewritten query; Confirm the fields to be collected based on user input, historical user conversations, and current categories; Complete field information extraction according to the fields that need to be collected, splice the extracted field information with the preset second prompt, and input it into the big model to output data query instructions; Obtain the query results of the field data to be collected through data query instructions; The query results are spliced ​​with the preset third prompt and input into the big model to rewrite the query results and generate content output.

6. The intelligent customer service response method based on a large model according to claim 5 is characterized in that: The method for obtaining the query result of the field data to be collected through the data query instruction includes: Update the extracted field information to the collection information field library; Whenever the collected information field library is updated, determine whether there are empty fields and confirm the number of questions corresponding to each empty field; When the number of questions about an empty field is less than two, select the corresponding empty fields in turn, concatenate the empty fields with the preset fourth prompt, and input the large model to output an empty field question instruction, ask questions to the user, and update the corresponding empty fields after receiving the reply; When there is no empty field or the number of questions about an empty field is greater than or equal to two, the confirmed field information is summarized; Perform data query on the offline database based on the summarized field information and obtain the query results.

7. The intelligent customer service response method based on a large model according to claim 6 is characterized in that: The method of performing data query on an offline database according to the aggregated field information and obtaining the query result includes: Determine whether the user is an account manager based on the current user's tag. If yes, then perform data query in the offline database according to the summarized field information and obtain the query result; Otherwise, the information of the current user is reviewed. If the user is confirmed to be the person with the current user tag, data query is performed on the offline database based on the summarized field information and the query results are obtained. Otherwise, it is transferred to manual.

8. An intelligent customer service response system based on a large model, characterized in that: include: The input module obtains user input and historical user conversations, concatenates them with the preset first prompt, and inputs them into the large model to output query rewriting and intent recognition; The processing module responds to the intention recognition, and splices the intention recognition with the preset second prompt, inputs the large model, and outputs the call instruction; The output module responds to the call instructions and generates content output.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method described in any one of claims 1 to 7 is implemented.

Images