Multi-domain knowledge fusion enhancement method based on large model capability advance

Abstract

The invention relates to a multi-domain knowledge fusion enhancement method based on large model ability advance, which comprises the following steps: constructing a question and answer library which comprises a plurality of pieces of data, and each piece of data is formed by combining a single question and an answer; when the large model receives a question input by a user, searching whether the question exists in the question answer library, and if the question is searched, returning a corresponding answer to the user; if the question is not retrieved, the large model generates an answer based on the question, returns the answer to the user, combines the question with the generated answer, and stores the combined question and answer in a question and answer library; the large model is provided with a question and answer library optimization mechanism, and a question and answer library is optimized through the question and answer library optimization mechanism; the large model is provided with a question self-generating mechanism, and questions and corresponding answers are randomly generated in a question answer library through the question self-generating mechanism.

Description

TECHNICAL FIELD

[0001] The application relates to a multi-field knowledge fusion enhancement method based on large model capability pre-branching, and belongs to the technical field of artificial intelligence. BACKGROUND

[0002] The model has strong understanding and reasoning capabilities, but in the actual application process, due to the lack of specific field professional knowledge, an external knowledge base is often used, such as the RAG method. In the power field, in the application process of a large model, multiple knowledge bases need to be hung, such as a device information base, operation inspection strategy knowledge, power supply service command procedure files and the like.

[0003] The existing RAG method and the like only call the large model once for the result of each business, which is equivalent to only being temporarily empowered by the large model once, and the precision is low, which cannot meet the application requirements. Mainly embodied in: Each time the business is applied, the user's question is mapped to multiple knowledge of multiple bases, and the required knowledge cannot be quickly and accurately queried; the multi-field knowledge retrieved cannot be quickly fused, and the multiple knowledge is only simply listed together, the quality of the prompt word of the large model is extremely low, and the capability of the large model cannot be exerted; the large model is only called by the user in the business application, and the result of each business is equivalent to only being temporarily empowered by the large model once, and the precision cannot be too high; the large model is in an idle state outside the time of user calling, and the reasoning capability of the large model is seriously wasted The above problems result in that the large model has good effects in general demonstration applications, but has poor effects in business landing applications. SUMMARY

[0004] In order to solve the problems in the prior art, the application provides a multi-field knowledge fusion enhancement method based on large model capability pre-branching.

[0005] The technical scheme of the application is as follows: On the one hand, the application provides a multi-field knowledge fusion enhancement method based on large model capability pre-branching, comprising the following steps: A question and answer database is constructed, and the question and answer database comprises multiple data, and each data is composed of a single question and an answer; When the large model receives a question input by a user, it is searched in the question and answer database whether the question exists, if the question is searched, the corresponding answer is returned to the user; If the question is not searched, the large model returns an answer based on the question to the user, and the question and the generated answer are combined and stored in the question and answer database; The large model is provided with a question and answer database optimization mechanism, and the question and answer database is optimized through the question and answer database optimization mechanism; The large model is provided with a question self-generation mechanism for randomly generating questions and corresponding answers in the question answer library.

[0006] Preferably, the large model is constructed based on a retrieval-enhanced generation technique, and the large model comprises a plurality of professional knowledge bases in different fields.

[0007] Preferably, the question self-generation mechanism specifically comprises: The large model randomly calls any professional knowledge base, randomly samples the content of the professional knowledge base to generate a plurality of questions and answers for each question, combines each question with the corresponding answer, and stores them in the question answer library.

[0008] Preferably, the question answer library optimization mechanism specifically comprises: The large model traverses each data in the question answer library one by one, evaluates the accuracy of the answer, and if the accuracy is lower than the preset accuracy threshold, the multi-field knowledge fusion large model generates a new answer based on the original answer and the knowledge base and replaces it to the original answer.

[0009] Preferably, the large model is provided with a priority mechanism, and the answer to the user question is set as the highest priority. Without interrupting the answer service of the user question, the question answer library optimization and the question self-generation are executed.

[0010] On the other hand, the present application also provides a multi-field knowledge fusion enhanced system based on large model capacity prepayment, comprising a question answer library construction module, a user business execution module, a question answer library optimization module and a question generation module. The question answer library construction module is used to construct a question answer library, and the question answer library comprises a plurality of data, each data being combined by a single question and an answer. The user business execution module is used to receive a question input by a user through a large model, search whether the question exists in the question answer library, return the corresponding answer to the user if the question is searched, return the answer generated by the large model to the user if the question is not searched, and store the question and the generated answer in the question answer library. The question answer library optimization module is used to set a question answer library optimization mechanism for optimizing the question answer library through the question answer library optimization mechanism. The question generation module is used to set a question self-generation mechanism for randomly generating questions and corresponding answers in the question answer library.

[0011] Preferably, the large model is constructed based on a retrieval-enhanced generation technique, and the large model comprises a plurality of professional knowledge bases in different fields.

[0012] Preferably, the large model is provided with a priority mechanism, and generation of answers to user questions is set as the highest priority, and the question answer library optimization and question self-generation are performed without interrupting the service of generating answers to user questions.

[0013] In another aspect, the present application also provides 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 method according to the present application when executing the program.

[0014] In another aspect, the present application also provides a computer readable storage medium having a computer program stored thereon, wherein the program is executable on a processor to implement the method according to the present application.

[0015] The present application has the following beneficial effects: 1. The present application fully integrates the knowledge in multiple service knowledge bases, utilizes the ability of the large model, repeatedly iterates and refines the answers to the questions, obtains an accurate question answer library, and does not need to output the large model every time the user has a service question, so that the ability of the large model is "prepaid" before the user calls the large model, and the accuracy of the answers to the questions is continuously improved through repeated prepayment, and the existing questions in the question answer library can be directly and quickly output, thereby effectively solving the problems that the large model cannot quickly and accurately query knowledge, the quality of the prompt words is low, the ability of the large model cannot be fully utilized, the running time of the large model is short and the ability is not fully utilized, and the performance is wasted due to the idle of the large model.

[0016] 2. The present application sets the large model service execution and the question answer library optimization mechanism in parallel, and based on the priority mechanism, the automatic question generation and knowledge base optimization in the background of the system can be realized without interrupting the user service, and the utilization rate of hardware resources (such as computing power and storage) is improved.

[0017] 3. Based on the random sampling of the professional knowledge base and the generation ability of the large model, the present application automatically generates possible high-frequency questions, generates question answer pairs in advance, fills the question answer library in advance, reduces the service peak pressure, continuously generates vertical field questions, quickly expands the applicability of the system in the subdivided scene, and reduces the cost of manual answer review. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The present application is a method flowchart. DETAILED DESCRIPTION

[0019] With reference to the drawings and the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by those skilled in the art without creative efforts belong to the scope of the present application.

[0020] It should be understood that the step numbers used herein are only for the convenience of description, and are not limited to the execution sequence of the steps.

[0021] It should be understood that the terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application and the appended claims, unless otherwise clear from the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0022] The terms "comprise" and "include" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0023] The term "and / or" means any combination of one or more of the associated listed items and all possible combinations thereof, and includes these combinations.

[0024] Referring to Figure 1 A multi-domain knowledge fusion enhancement method based on large model capacity pre-branching, comprising the following steps: A question and answer library is constructed, which includes a plurality of data, each data being composed of a single question and an answer; When the large model receives a question input by a user, it searches the question and answer library to determine whether the question exists, and if the question is found, the corresponding answer is returned to the user; If the question is not found, the large model generates an answer based on the question and returns it to the user, and then combines the question with the generated answer and stores them in the question and answer library; The large model is provided with a question and answer library optimization mechanism to optimize the question and answer library; The large model is provided with a question self-generation mechanism to randomly generate questions and corresponding answers in the question and answer library.

[0025] In a specific embodiment, the large model business execution, the question and answer library optimization mechanism, and the question self-generation mechanism are run in parallel; In one specific embodiment, the question self-generation mechanism is triggered periodically. For example, based on a professional medical knowledge base, the large model can randomly generate several questions, such as: Question 1: "What are the common hazards of high blood pressure?" Answer 1: "High blood pressure can cause heart disease, stroke, and kidney damage, etc." Question 2: "What are the control methods for diabetes?" Answer 2: "Including diet control, regular exercise, drug treatment and blood glucose monitoring." The generated questions and answers are automatically written into the question and answer library, thereby expanding the library capacity and improving coverage.

[0026] In some embodiments, the large model is constructed based on retrieval enhancement generation technology, and the large model includes multiple professional knowledge bases in different fields.

[0027] In one specific embodiment, the large model retrieves in different professional knowledge bases based on BAAI or bge-m3 model.

[0028] In one specific embodiment, the retrieval engine is Faiss.

[0029] In one specific embodiment, the question and answer library is constructed based on Redis.

[0030] In one specific embodiment, the large model is constructed based on Deepseek-R1 model or Deepseek-V3 model.

[0031] In some embodiments, the question self-generation mechanism is specifically: The large model randomly calls any professional knowledge base, randomly samples the content of the professional knowledge base to generate multiple questions and answers for each question, and then combines each question with the corresponding answer and stores it in the question and answer library.

[0032] In some embodiments, the question and answer library optimization mechanism is specifically: The large model traverses each data in the question and answer library one by one, and evaluates the accuracy of the answer. If the accuracy is lower than the preset accuracy threshold, the multi-field knowledge fusion large model generates a new answer based on the original answer and the knowledge base and replaces it to the original answer. For example, if the answer to the question "What are the main applications of blockchain?" is "mainly used for digital currency", and the accuracy is lower than the preset threshold (such as 70%), the system will regenerate the answer based on multi-field knowledge fusion: New answer: "Blockchain is not only used for digital currency, but also widely used in supply chain management, digital identity authentication, smart contract and data traceability, etc."

[0033] In some embodiments, the large model is provided with a priority mechanism, which sets the generation of answers to user questions as the highest priority, and executes the optimization of the question answer library and the self-generation of questions without interrupting the service of generating answers to user questions. For example, when the user is consulting "what are the main challenges of photovoltaic power generation", even if the system is executing a self-generation task in the background, the background process will be immediately suspended, and all computing resources will be used to quickly return the answer to the user's question, thereby guaranteeing the service experience.

[0034] In some embodiments, a capability enhancement system for a large model is proposed, which includes a question answer library construction module, a user business execution module, a question answer library optimization module, and a question generation module. The question answer library construction module is used to construct a question answer library, which includes a plurality of data, each data being composed of a single question and an answer. The user business execution module is used to receive a user input question through the large model, search for whether the question exists in the question answer library, return the corresponding answer to the user if the question is searched, and return the answer generated by the large model to the user if the question is not searched, and combine the question and the generated answer and store them in the question answer library. The question answer library optimization module is used to set a question answer library optimization mechanism to optimize the question answer library through the question answer library optimization mechanism. The question generation module is used to set a question self-generation mechanism to randomly generate questions and corresponding answers in the question answer library through the question self-generation mechanism.

[0035] In some embodiments, the large model is constructed based on a retrieval enhancement generation technology, and the large model includes a plurality of professional knowledge bases in different fields.

[0036] In some embodiments, the large model is provided with a priority mechanism, which sets the generation of answers to user questions as the highest priority, and executes the optimization of the question answer library and the self-generation of questions without interrupting the service of generating answers to user questions.

[0037] In some embodiments, an electronic device is proposed, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the method according to any one of the embodiments of the present application when executing the program.

[0038] In some embodiments, a computer readable storage medium is proposed, which stores a computer program executable by a processor, and the program implements the method according to any one of the embodiments of the present application when executed by the processor.

[0039] In the embodiments of the present application, “at least one” means one or more, and “multiple” means two or more. “And / or” describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent the cases of A alone, A and B together, and B alone. Wherein A and B can be singular or plural. The character “ / ” generally represents an “or” relationship between the front and rear associated objects. “At least one of the following” and the like means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0040] Those of ordinary skill in the art can realize that the units and algorithm steps described in the embodiments disclosed herein can be realized in electronic hardware, computer software, and a combination of electronic hardware and computer software. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0041] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0042] In several embodiments provided by the present application, any function realized in the form of a software function unit and sold or used as an independent product can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-Only Memory; hereinafter referred to as: ROM), a random access memory (Random Access Memory; hereinafter referred to as: RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0043] The above is only an embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation based on the content of the specification and drawings of the present application, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A multi-domain knowledge fusion enhancement method based on large model capability pre-support, characterized in that, Includes the following steps: Construct a question-and-answer database, which includes multiple data entries, each consisting of a single question and answer. When the large model receives a question input by the user, it searches the question-answer database to see if the question already exists. If the question is found, the corresponding answer is returned to the user. If the question is not found, the large model generates an answer based on the question and returns it to the user. The question and the generated answer are then combined and stored in the question-answer database. The large model is equipped with a question-answer database optimization mechanism to optimize the question-answer database. The large model has a question self-generation mechanism, which randomly generates questions and corresponding answers from the question and answer database.

2. The multi-domain knowledge fusion enhancement method based on large model capability pre-support as described in claim 1, characterized in that, The large model is built based on retrieval-enhanced generation technology and includes professional knowledge bases from multiple different fields.

3. The multi-domain knowledge fusion enhancement method based on large model capability pre-support as described in claim 2, characterized in that, The problem self-generation mechanism is as follows: The large model randomly calls any professional knowledge base, randomly samples the content of the professional knowledge base to generate multiple questions and generates answers for each question, and then combines each question with its corresponding answer and stores them in the question-answer database.

4. The multi-domain knowledge fusion enhancement method based on large model capability pre-support as described in claim 3, characterized in that, The optimization mechanism for the question answer database is as follows: The large model iterates through each piece of data in the question-answer database and evaluates the accuracy of the answer. If the accuracy is lower than the preset accuracy threshold, the multi-domain knowledge fusion large model regenerates a new answer based on the original answer and the knowledge base and replaces the original answer.

5. The multi-domain knowledge fusion enhancement method based on large model capability pre-support as described in claim 4, characterized in that, The large model has a priority mechanism that sets the generation of answers to user questions as the highest priority, and performs question answer database optimization and question self-generation without interrupting the service of generating answers to user questions.

6. A multi-domain knowledge fusion enhancement system based on large model capability pre-support, characterized in that, It includes a question-and-answer database construction module, a user business execution module, a question-and-answer database optimization module, and a question generation module; The question-answer database construction module is used to construct a question-answer database, which includes multiple data entries, each of which consists of a single question and answer. The user service execution module is used to receive questions input by users through a large model, search the question answer database to see if the question already exists, and if the question is found, return the corresponding answer to the user. If the question is not found, the large model generates an answer based on the question and returns it to the user. The question and the generated answer are then combined and stored in the question-answer database. The question-answer database optimization module is used to set up a question-answer database optimization mechanism to optimize the question-answer database. The question generation module is used to set up a question self-generation mechanism, which randomly generates questions and corresponding answers from the question and answer database.

7. A multi-domain knowledge fusion enhancement system based on large model capability pre-support as described in claim 6, characterized in that, The large model is built based on retrieval-enhanced generation technology and includes professional knowledge bases from multiple different fields.

8. The multi-domain knowledge fusion enhancement system based on large model capability pre-support as described in claim 7, characterized in that, The large model has a priority mechanism that sets the generation of answers to user questions as the highest priority, and performs question answer database optimization and question self-generation without interrupting the service of generating answers to user questions.

9. 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, it implements the method as described in any one of claims 1 to 5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 5.

Images