Business Q&A Method and System Based on Large Language Model and Scenario Labels

Abstract

The present invention belongs to the technical field of intelligent question answering, and specifically discloses a business question answering method and system based on a large language model and scenario tags. Through business question expansion and matching, the corresponding knowledge base question set is associated, and then the answer is retrieved from the knowledge base to determine the corresponding data block and scenario tag. Then, through scenario tag matching and scenario consistency evaluation, a database with high scenario consistency and the user's question combination are selected and input into the large language model to achieve intelligent question answering. It can effectively solve the user's business question answering needs without the user directly querying professional knowledge; it can analyze the questions raised by the user more comprehensively, fully covering the question intention and question scenario information, making the question answering feedback result more targeted and more relevant to the user's business questions; and it can make the question answering result more concise and effective.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent question answering, and particularly relates to a business question answering method and system based on a large language model and scenario tags. Background Art

[0002] An intelligent question answering system sorts out the accumulated disordered corpus information in an orderly and scientific manner and establishes a knowledge-based classification model. These classification models can guide newly added corpus consultation and service information, saving human resources and improving the automation of information processing. The intelligent question answering system, in the form of question and answer, accurately locates the question knowledge required by the user and provides personalized information services to the user through interaction. Existing intelligent question answering systems mainly have three implementation methods:

[0003] One is the intelligent question answering technology mainly based on traditional search technology. According to the string matching situation or semantic similarity of the question input by the user, a "question-answer" match is directly performed from the knowledge base.

[0004] Two is the intelligent question answering technology oriented to knowledge graphs. The knowledge graph is the main form of the knowledge base, and the intelligent question answering technology system converts the natural language input conditions of the user into query statements and inference formulas for the graph database.

[0005] Three is the intelligent question answering technology based on large language models. Using the Retrieval-Augmented Generation (RAG) technology framework, first, data blocks semantically related to the question input by the user are queried from the vector database. Then, the data blocks are spliced into the prompt input conditions of the large language model. Finally, the large language model combines the question and the content of the data blocks to give a dialogue answer feedback.

[0006] Existing intelligent question answering technologies mainly still have the following defects:

[0007] 1. The intelligent question answering technology mainly based on traditional search technology has a single expression form, a limited range of question answering capabilities, and a low question matching degree.

[0008] 2. The intelligent question answering technology based on knowledge graphs mainly solves the requirements of factual questions and is difficult to solve personalized question answering requirements. The knowledge graph has limited expression ability for complex knowledge.

[0009] 3. The intelligent question answering technology based on large language models has a flexible expression form and can give a certain reasonable reply whether or not relevant data blocks are matched. However, the answer result of the large language model largely depends on the query effect of the data blocks. Querying data blocks directly according to the user's question, the obtained data results have a low relevance to the user's question requirements and cannot comprehensively cover the scenario information of the user's question. Therefore, the answer content is too redundant or lacks pertinence.

[0010] 4 Existing intelligent Q&A systems mainly adopt a technical architecture targeting knowledge-based question requirements. In actual business scenarios, users generally do not directly ask questions about professional knowledge (such as "How to apply for a publication business license?", "What industrial and commercial qualifications should be possessed for selling books?", "What are the requirements for the business premises for selling books offline?", etc.), but ask questions in the form of business requirements (such as "How to operate a bookstore?"). For business questions, the logical compatibility of existing Q&A technologies is poor, and the accuracy of returned results is insufficient. Summary of the Invention

[0011] The object of the present invention is to provide a business Q&A method and system based on a large language model and scenario tags to solve the above problems existing in the prior art.

[0012] To achieve the above object, the present invention adopts the following technical solutions:

[0013] In the first aspect, a business Q&A method based on a large language model and scenario tags is provided, including:

[0014] Obtain the business question input by the user and the supplementary description of the business question;

[0015] Import the business question into a pre-set large language model for question expansion to obtain the corresponding expanded question information;

[0016] Based on the expanded question information, match and determine the corresponding associated question set in the knowledge base, and the associated question set includes several knowledge base associated questions;

[0017] In response to the user's operation instruction, select an associated question subset including several question items from the associated question set, and retrieve the knowledge base associated answers corresponding to each question item in the associated question subset from the knowledge base as the question reference answers for the corresponding associated question subset;

[0018] Determine the data blocks included in each question reference answer in the knowledge base and the set of scenario tags associated with the data blocks;

[0019] For each question item of the associated question subset, perform a scenario consistency evaluation on the set of scenario tags of each corresponding data block and the supplementary description of the business question to obtain the scenario consistency scores of each data block;

[0020] Select several data blocks with the highest scenario consistency scores as the target data blocks for each question item of the associated question subset, and combine the target data blocks corresponding to each question item of the associated question subset, the business question, and the supplementary description of the business question and import them into a pre-set large language model for natural language processing to obtain the corresponding business answers respectively, and finally fuse multiple business answers to obtain a unified answer output result.

[0021] In a possible design, matching and determining a corresponding associated question set in the knowledge base based on the expanded question information includes:

[0022] Traverse the question list in the knowledge base, where the question list contains several candidate knowledge base associated questions;

[0023] If the trigger word label of a certain candidate knowledge base associated question in the question list is in the expanded question information, then include the candidate knowledge base associated question in the associated question set. In the knowledge base, each candidate knowledge base associated question contains one or more trigger word labels, which are confirmed in the process of building the knowledge base. When entering each candidate knowledge base associated question, the system automatically identifies 1 to N candidate trigger words through word segmentation, and is selected and confirmed by the user or supplemented and confirmed manually to obtain the corresponding trigger word label.

[0024] In a possible design, selecting an associated question subset from the associated question set in response to a user's operation instruction includes:

[0025] Perform semantic similarity matching between each knowledge base associated question in the associated question set and the business question, and sort each knowledge base associated question according to the semantic similarity between each knowledge base associated question and the business question to obtain a sorted associated question set;

[0026] Output and display the sorted associated question set to the user, and select several knowledge base associated questions from the sorted associated question set to form an associated question subset in response to the user's operation instruction.

[0027] In a possible design, determining each data block included in the question reference answer and the set of scenario labels associated with each data block includes:

[0028] Determine each data block included in the question reference answer and several scenario labels associated with each data block;

[0029] Perform a summary process on the scenario labels associated with each data block to obtain the set of scenario labels for each data block.

[0030] In a possible design, performing a scenario consistency evaluation on the set of scenario labels of each data block and the supplementary description of the business question to obtain the scenario consistency score of each data block includes:

[0031] Traverse the set of scenario labels of each data block, and determine whether each scenario label in its set of scenario labels is included in the supplementary description of the business question;

[0032] If a certain scenario label in the set of scenario labels of a data block is included in the supplementary description of the business question, then add one point to the scenario consistency score of that data block;

[0033] Statistically obtain the final scene consistency scores for each data block.

[0034] In a possible design, if a certain scene label in the scene label set of a data block is not included in the supplementary description of the business problem, then determine the first semantic vector of the corresponding scene label and the second semantic vector of the supplementary description, substitute the first semantic vector of the corresponding scene label and the second semantic vector of the supplementary description into a preset semantic similarity scoring formula for calculation to obtain the corresponding semantic similarity score, and count the semantic similarity score into the scene consistency score of the corresponding data block. The semantic similarity scoring formula is:

[0035]

[0036] where Score(d, t i ) represents the semantic similarity score between the corresponding scene label and the supplementary description, represents the first semantic vector of the corresponding scene label, v d represents the second semantic vector of the supplementary description.

[0037] In a possible design, selecting several data blocks with the highest scene consistency scores as target data blocks includes: selecting three data blocks with the highest scene consistency scores as target data blocks.

[0038] In a second aspect, a business Q&A system based on a large language model and scene labels is provided, including a question acquisition unit, a question expansion unit, an association matching unit, a knowledge screening unit, a scene annotation unit, a scene matching unit, and an intelligent Q&A unit, where:

[0039] The question acquisition unit is used to acquire the business question input by the user and the supplementary description of the business question;

[0040] The question expansion unit is used to import the business question into a preset large language model for question expansion to obtain the corresponding expanded question information;

[0041] The association matching unit is used to match and determine the corresponding associated question set in the knowledge base based on the expanded question information. The associated question set contains several knowledge base associated questions;

[0042] The knowledge screening unit is used to select an associated question subset containing several question items from the associated question set in response to the user's operation instruction, and retrieve the knowledge base associated answers corresponding to each question item in the associated question subset from the knowledge base as the question reference answers for the corresponding associated question subset;

[0043] A scenario annotation unit, configured to determine data blocks included in the reference answer of each question in the knowledge base, and a set of scenario labels associated with each data block;

[0044] A scenario matching unit, configured to, for each question item in the associated question subset, perform a scenario consistency evaluation on the set of scenario labels of each corresponding data block and the supplementary description of the business question, to obtain the scenario consistency score of each data block;

[0045] An intelligent question-answering unit, configured to select several data blocks with the highest scenario consistency scores as the target data blocks for each question item in the associated question subset, and combine the target data blocks corresponding to each question item in the associated question subset, the business question, and the supplementary description of the business question, and import them into a pre-set large language model for natural language processing, to obtain corresponding business answers respectively, and finally fuse multiple business answers to obtain a unified answer output result.

[0046] In a third aspect, there is provided a business question-answering system based on a large language model and scenario labels, including:

[0047] A memory, configured to store instructions;

[0048] A processor, configured to read the instructions stored in the memory and execute any one of the methods in the first aspect according to the instructions.

[0049] In a fourth aspect, there is provided a computer-readable storage medium, on which instructions are stored, and when the instructions run on a computer, the computer is caused to execute any one of the methods in the first aspect. At the same time, there is also provided a computer program product, and when the computer program product runs on a computer, it executes any one of the methods in the first aspect.

[0050] Advantageous effects: The present invention can effectively solve the business question-answering needs of users without the users directly querying professional knowledge; it can perform a more comprehensive analysis of the questions raised by users, fully covering the question intention and question scenario information; it makes the question-answering feedback result more relevant to the business questions of users, improving the interaction experience of users; and, through scenario label matching, the reach rate of business question-answering results can be improved, making the result content more targeted; at the same time, it can solve the knowledge base query task for long question-answering content, making the question-answering results more concise and effective. Description of the Drawings

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0052] Figure 1 It is a schematic diagram of the steps of the method in Embodiment 1 of the present invention;

[0053] Figure 2 It is a schematic diagram of the composition of the system in Embodiment 2 of the present invention;

[0054] Figure 3 It is a schematic diagram of the composition of the system in Embodiment 3 of the present invention. Detailed implementation manners

[0055] It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation on the present invention. The specific structural and functional details disclosed herein are only used to describe the exemplary embodiments of the present invention. However, the present invention can be embodied in many alternative forms and should not be construed as limited to the embodiments set forth herein.

[0056] It should be understood that unless otherwise clearly defined and limited, the term "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meaning of the above terms in the embodiments can be understood according to specific circumstances.

[0057] Specific details are provided in the following description to facilitate a complete understanding of the exemplary embodiments. However, those of ordinary skill in the art should understand that the exemplary embodiments can be implemented without these specific details. For example, the device can be shown in a block diagram to avoid making the example unclear with unnecessary details. In other embodiments, well-known processes, structures, and technologies can be shown without unnecessary details to avoid making the embodiments unclear.

[0058] Embodiment 1:

[0059] This embodiment provides a business Q&A method based on a large language model and scenario tags, which can be applied to a corresponding Q&A server, such as Figure 1 shown, the method includes the following steps:

[0060] S1. Obtain the business question input by the user and the supplementary description of the business question.

[0061] In specific implementation, the user can input a corresponding business question and supplementary instructions for the business question to the Q&A server. The supplementary instructions are relevant description / interpretation text content for the business question, including relevant scenario / background limitation information of the question, etc. After the Q&A server obtains the business question and supplementary instructions input by the user, it performs subsequent analysis and processing.

[0062] S2. Import the business question into a pre-set large language model for question expansion to obtain corresponding expanded question information.

[0063] In specific implementation, the Q&A server can import the business question into a pre-set large language model for question expansion query to obtain the expanded question information of the business question. By expanding and strengthening the business question, more question clues can be provided to the knowledge Q&A engine of the model for subsequent use.

[0064] S3. Based on the expanded question information, match and determine a corresponding set of associated questions in the knowledge base. The set of associated questions contains several knowledge base associated questions.

[0065] In specific implementation, the Q&A server traverses to determine the trigger word tags (i.e., corresponding keywords) included in each expanded question in the expanded question information, and matches and determines the knowledge base associated questions corresponding to each trigger word tag in the knowledge base. Then, the knowledge base associated questions corresponding to each trigger word tag are used to form a set of associated questions. A number of candidate knowledge base associated questions and knowledge base associated answers corresponding to each candidate knowledge base associated question are pre-configured in the knowledge base, and each knowledge base associated question is associated with a corresponding trigger word tag (i.e., corresponding keyword). In the knowledge base, each candidate knowledge base associated question contains one or more trigger word tags. The trigger word tags are confirmed in the process of constructing the knowledge base. When entering each candidate knowledge base associated question, the system automatically determines 1 to N candidate trigger words through word segmentation, which are selected and confirmed by the user or supplemented and confirmed manually to obtain the corresponding trigger word tags. The trigger word tags are on the question, and the subsequent scenario tags are on the data block of the question answer.

[0066] S4. In response to the user's operation instruction, select a subset of associated questions containing several question items from the set of associated questions, and retrieve the knowledge base associated answers corresponding to each question item in the subset of associated questions from the knowledge base as the question reference answers for the corresponding subset of associated questions.

[0067] In specific implementation, the Q&A server may perform semantic similarity matching between each knowledge base associated question in the set of associated questions and the business question, and sort each knowledge base associated question according to the semantic similarity between the knowledge base associated question and the business question, so as to obtain a sorted set of associated questions. Then, the sorted set of associated questions is output and displayed to the user, so that the user can view, through a human-computer interaction method, which knowledge base associated questions match the knowledge questions of concern and make selections for the corresponding knowledge base associated questions. The Q&A server may, in response to the user's operation instruction, select several knowledge base associated questions from the sorted set of associated questions to form an associated question subset, and retrieve, from the knowledge base, the knowledge base associated answers corresponding to each knowledge base associated question in the associated question subset as the question reference answers.

[0068] S5. Determine the data blocks included in each question reference answer in the knowledge base, and the set of scenario tags associated with the data blocks.

[0069] In specific implementation, the Q&A server first determines each data block included in the question reference answer and several scenario tags associated with each data block. The question reference answer usually has the content structure of a long answer. Each question reference answer includes at least 1 data block, and each data block is associated with multiple scenario tags (i.e., corresponding scenario keywords). According to the number of scenario tags, a nested content structure is formed. In actual business processes, the reference to knowledge such as operation instructions, method suggestions, management systems, and regulations is closely related to specific sub-scenario tags. For example, if the corresponding question 1 in the knowledge base is recorded as Q1, and the question reference answer 1 is divided into multiple data block parts according to paragraphs, and each data block part is composed of several scenario tags. The corresponding format example is as follows:

[0070] Q1:

[0071] A1-1: {T1}

[0072] A1-2: {T1, T2, T3}

[0073] A1-3: {T1, T2, T4}

[0074] A1-4: {T1, T5}

[0075] A1-5: {T1, T5, T6}

[0076] The data blocks included in the question reference answer 1 are A1-1, A1-2, A1-3, A1-4, and A1-5, and each data block is associated with multiple scenario tags such as T1, T2, T3, etc. The scenario tags associated with each data block can be summarized to obtain the set of scenario tags for each data block, such as {T1, T2, T3}, etc.

[0077] S6. For each problem item in the subset of associated problems, perform a scenario consistency assessment on the set of scenario tags of each corresponding data block and the supplementary description of the business problem to obtain the scenario consistency scores of each data block.

[0078] Specifically, when implemented, the Q&A server can traverse the set of scenario tags of each data block and determine whether each scenario tag (i.e., the corresponding scenario keyword) in its set of scenario tags is included in the supplementary description of the business problem. If a certain scenario tag in the set of scenario tags of a data block is included in the supplementary description of the business problem, the scenario consistency score of that data block is incremented by one point, and then the final scenario consistency scores of each data block are statistically obtained. If a certain scenario tag in the set of scenario tags of a data block is not included in the supplementary description of the business problem, the first semantic vector of the corresponding scenario tag and the second semantic vector of the supplementary description are determined, and the first semantic vector of the corresponding scenario tag and the second semantic vector of the supplementary description are substituted into a preset semantic similarity scoring formula for calculation (taking cosine similarity as an example, it can also be replaced by any similarity metric function with a value in [0, 1]. For example, the semantic embedding models that can be used include text-embedding-ada-002 of OpenAI, embedding text2vec-base-chinese, BERT series models, FastText, etc.) to obtain the corresponding semantic similarity score, and the semantic similarity score is statistically included in the scenario consistency score of the corresponding data block, and then the final scenario consistency scores of each data block are statistically obtained. The semantic similarity scoring formula is:

[0079]

[0080] where Score(d, t i ) represents the semantic similarity score between the corresponding scenario tag and the supplementary description, represents the first semantic vector of the corresponding scenario tag, and v d represents the second semantic vector of the supplementary description.

[0081] S7. Select several data blocks with the highest scenario consistency scores as the target data blocks for each problem item in the subset of associated problems, and combine the target data blocks corresponding to each problem item in the subset of associated problems, the business problem, and the supplementary description of the business problem and import them into a preset large language model for natural language processing to obtain the corresponding business answers respectively, and finally fuse multiple business answers to obtain a unified answer output result.

[0082] During specific implementation, the Q&A server selects several data blocks with the highest scene consistency scores. For example, the three data blocks with the highest scene consistency scores are used as the target data blocks for each question item of the corresponding associated question subset. Then, the target data blocks corresponding to each question item of the associated question subset, the business question, and the supplementary description of the business question are combined and imported into a pre-set large language model for natural language processing to obtain corresponding business answers respectively. Finally, multiple business answers are fused to obtain a unified answer output result.

[0083] The method of this embodiment can effectively solve the user's business Q&A needs without the user directly querying professional knowledge; it can analyze the questions raised by the user more comprehensively, fully covering the question intention and question scenario information; it makes the Q&A feedback result more relevant to the user's business question, improving the user's interaction experience; and, through scene label matching, the reach rate of the business Q&A result can be increased, making the result content more targeted; at the same time, it can solve the knowledge base query task for long Q&A content, making the Q&A result more concise and effective.

[0084] Embodiment 2:

[0085] This embodiment provides a business Q&A system based on a large language model and scene labels, as Figure 2 shown, including a question acquisition unit, a question expansion unit, an association matching unit, a knowledge screening unit, a scene annotation unit, a scene matching unit, and an intelligent Q&A unit, where:

[0086] The question acquisition unit is used to acquire the business question input by the user and the supplementary description of the business question;

[0087] The question expansion unit is used to import the business question into a pre-set large language model for question expansion to obtain the corresponding expanded question information;

[0088] The association matching unit is used to match and determine the corresponding associated question set in the knowledge base based on the expanded question information, and the associated question set contains several knowledge base associated questions;

[0089] The knowledge screening unit is used to select an associated question subset containing several question items from the associated question set in response to the user's operation instruction, and retrieve the knowledge base associated answers corresponding to each question item in the associated question subset from the knowledge base as the question reference answers for the corresponding associated question subset;

[0090] The scene annotation unit is used to determine the data blocks included in each question reference answer in the knowledge base, and the set of scene labels associated with each data block;

[0091] A scene matching unit, which is used to evaluate the scene consistency between the set of scene labels of each data block corresponding to each problem item in the associated problem subset and the supplementary description of the business problem, and obtain the scene consistency scores of each data block;

[0092] An intelligent question-answering unit, which is used to select several data blocks with the highest scene consistency scores as the target data blocks for each problem item in the associated problem subset, and combine the target data blocks, business problems, and supplementary descriptions of the business problems corresponding to each problem item in the associated problem subset and import them into a pre-set large language model for natural language processing to obtain corresponding business answers respectively, and finally fuse multiple business answers to obtain a unified answer output result.

[0093] Embodiment 3:

[0094] This embodiment provides a business Q&A system based on a large language model and scene labels, as Figure 3 shown. At the hardware level, it includes:

[0095] A data interface, which is used to establish data docking between the processor and an external data terminal;

[0096] A memory, which is used to store instructions;

[0097] A processor, which is used to read the instructions stored in the memory and execute the business Q&A method based on the large language model and scene labels in Embodiment 1 according to the instructions.

[0098] Optionally, the system further includes an internal bus. The processor, the memory, and the data interface can be interconnected through the internal bus. The internal bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0099] The memory may but is not limited to include Random Access Memory (RAM), Read Only Memory (ROM), Flash Memory, First Input First Output (FIFO), and / or First In Last Out (FILO), etc. The processor may be a general-purpose processor, including Central Processing Unit (CPU), Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0100] Embodiment 4:

[0101] This embodiment provides a computer-readable storage medium, on which instructions are stored. When the instructions run on a computer, the computer is caused to execute the business question-answering method based on the large language model and scenario tags in Embodiment 1. Among them, the computer-readable storage medium refers to a carrier for storing data, which may but is not limited to include floppy disks, optical discs, hard disks, flash memories, USB flash drives, and / or Memory Sticks, etc. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0102] This embodiment also provides a computer program product, which when running on a computer, executes the business question-answering method based on the large language model and scenario tags in Embodiment 1. Among them, the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0103] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A business Q&A method based on large language models and scenario tags, characterized in that, including: Obtain the business problem input by the user and supplementary explanations for the business problem; Import the business problem into a pre - set large - language model for problem expansion to obtain corresponding expanded problem information; Based on the expanded problem information, match and determine the corresponding associated problem set in the knowledge base. The associated problem set contains several knowledge - base associated problems; In response to the user's operation instruction, select an associated problem subset containing several problem items from the associated problem set, and retrieve the knowledge - base associated answers corresponding to each problem item in the associated problem subset from the knowledge base as the problem reference answers for the corresponding associated problem subset; Determine the data blocks included in each problem reference answer in the knowledge base and the set of scenario tags associated with the data blocks; For each problem item in the associated problem subset, perform a scenario consistency evaluation on the set of scenario tags of its corresponding data blocks and the supplementary explanation of the business problem to obtain the scenario consistency scores of each data block; Select several data blocks with the highest scenario consistency scores as the target data blocks for each problem item in the associated problem subset, and combine the target data blocks corresponding to each problem item in the associated problem subset, the business problem, and the supplementary explanation of the business problem and import them into a pre - set large - language model for natural language processing to obtain corresponding business answers respectively. Finally, fuse multiple business answers to obtain a unified answer output result.

2. The business Q&A method based on the large language model and scenario tags according to claim 1, wherein The matching and determining the corresponding associated problem set in the knowledge base based on the expanded problem information includes: Traverse the problem list in the knowledge base. The problem list contains several candidate knowledge - base associated problems; If the trigger - word tag of a certain candidate knowledge - base associated problem in the problem list is in the expanded problem information, then include the candidate knowledge - base associated problem in the associated problem set.

3. The business Q&A method based on the large language model and scenario tags according to claim 1, wherein, The selecting the associated problem subset from the associated problem set in response to the user's operation instruction includes: Perform semantic similarity matching between each knowledge - base associated problem in the associated problem set and the business problem, and sort each knowledge - base associated problem according to the semantic similarity between each knowledge - base associated problem and the business problem to obtain a sorted associated problem set; Output and display the sorted associated problem set to the user, and in response to the user's operation instruction, select several knowledge - base associated problems from the sorted associated problem set to form an associated problem subset.

4. The business Q&A method based on a large language model and scenario tags according to claim 1, wherein The determining the data blocks included in each problem reference answer and the set of scenario tags associated with each data block includes: Determine the data blocks included in each problem reference answer and several scenario tags associated with each data block; Perform a summary process on the scenario tags associated with each data block to obtain the set of scenario tags for each data block.

5. The business Q&A method based on large language models and scenario tags according to claim 1, wherein The performing a scenario consistency evaluation on the set of scenario tags of each data block and the supplementary explanation of the business problem to obtain the scenario consistency scores of each data block includes: Traverse the set of scenario tags of each data block and determine whether each scenario tag in its set of scenario tags is included in the supplementary explanation of the business problem; If a certain scenario tag in the set of scenario tags of a data block is included in the supplementary explanation of the business problem, then add one point to the scenario consistency score of the data block; Statistically obtain the final scenario consistency scores of each data block.

6. The business Q&A method based on the large language model and scenario tags according to claim 5, characterized in that, If a certain scenario tag in the set of scenario tags of a data block is not included in the supplementary description of the business problem, determine the first semantic vector of the corresponding scenario tag and the second semantic vector of the supplementary description, substitute the first semantic vector of the corresponding scenario tag and the second semantic vector of the supplementary description into a preset semantic similarity scoring formula for calculation, obtain the corresponding semantic similarity score, and count the semantic similarity score into the scenario consistency score of the corresponding data block. The semantic similarity scoring formula is as follows: Among them, The first semantic vector representing the corresponding scene label, v d The second semantic vector representing supplementary description.

7. The business Q&A method based on the large language model and scenario tags according to claim 1, characterized in that The step of selecting several data blocks with the highest scenario consistency scores as target data blocks includes: selecting three data blocks with the highest scenario consistency scores as target data blocks.

8. A business Q&A system based on a large language model and scenario tags, characterized in that, It includes a problem acquisition unit, a problem expansion unit, an association matching unit, a knowledge screening unit, a scenario annotation unit, a scenario matching unit, and an intelligent question answering unit, where: The problem acquisition unit is used to obtain the business problem input by the user and the supplementary description of the business problem; The problem expansion unit is used to import the business problem into a preset large language model for problem expansion to obtain the corresponding expanded problem information; The association matching unit is used to match and determine the corresponding set of associated problems in the knowledge base based on the expanded problem information. The set of associated problems contains several knowledge base associated problems; The knowledge screening unit is used to select an associated problem subset containing several problem items from the set of associated problems in response to the user's operation instruction, and retrieve the knowledge base associated answers corresponding to each problem item in the associated problem subset from the knowledge base as the problem reference answers for the corresponding associated problem subset; The scenario annotation unit is used to determine the data blocks included in each problem reference answer in the knowledge base and the set of scenario tags associated with each data block; The scenario matching unit is used to perform a scenario consistency evaluation on the set of scenario tags of each data block corresponding to each problem item in the associated problem subset and the supplementary description of the business problem to obtain the scenario consistency scores of each data block; The intelligent question answering unit is used to select several data blocks with the highest scenario consistency scores as the target data blocks for each problem item in the associated problem subset, and combine the target data blocks corresponding to each problem item in the associated problem subset, the business problem, and the supplementary description of the business problem and import them into a preset large language model for natural language processing to obtain the corresponding business answers respectively, and finally fuse multiple business answers to obtain a unified answer output result.

9. A business Q&A device based on a large language model and scenario tags, characterized in that, It includes: A memory for storing instructions; A processor for reading the instructions stored in the memory and executing the business question answering method based on a large language model and scenario tags according to any one of claims 1-7.

Images