Retrieval enhancement generation-based personalized question and answer method and system for large language model

Abstract

The invention discloses a retrieval enhancement generation-based big language model personalized question and answer method, which comprises the following steps of: receiving a dialogue request of a client, and analyzing the dialogue request into personal condition description and a specific question; sending the specific question to a pre-constructed user portrait knowledge retrieval enhancement database, obtaining retrieval enhancement data containing a user portrait, and combining the retrieval enhancement data with the personal condition description into context retrieval enhancement data containing the user portrait; and according to the context retrieval enhancement data containing the user portrait and the specific question, generating a question instruction and inputting the question instruction into a large language model to obtain corresponding answer feedback data. According to the method, individual condition description and specific questions are analyzed, retrieval enhancement is carried out through the user portrait knowledge retrieval enhancement database, personalized adaptation of large language model questions and answers can be achieved, and the problem that in the prior art, the personalized degree of a large language model is not high is solved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method for designing a personalized question-answering method based on a large language model generated by retrieval enhancement. Background Art

[0002] With the advancement of artificial intelligence and natural language processing technologies, generative large language models (LLMs) have garnered widespread attention. LLMs are self-supervised and trained on massive amounts of text data, typically with billions of parameters. They are capable of performing natural language processing tasks such as text classification, question-answering, conversational speech, and sentiment analysis. Retrieval Augmented Generation (RAG), a widely adopted LLM application, enables a model to respond to user queries based on a specified set of documents. This information is then used to augment the model's already extensive static training data, resulting in responses that better meet user needs.

[0003] Currently, large language models continue to develop in the field of question-and-answer dialogue, but they have significant limitations when it comes to providing personalized responses to user needs. Furthermore, dialogue systems cannot dynamically adjust to changes in user emotions and preferences. Inaccurate or even contradictory judgments about user response tendencies can result in incorrect responses and even hurt user feelings. To further enhance the personalized needs of dialogue systems, existing dialogue systems often provide users with long-context conversation information in advance and then combine it with LLM to generate natural language responses. This allows the LLM in the dialogue system to infer user preferences from historical conversation information. However, existing dialogue systems still have some problems: (1) Low relevance of user profiles: Historical conversation information often contains a large amount of irrelevant information, making it difficult to infer user preferences. At the same time, most user information cannot be inferred from historical contextual information, resulting in failure in inferring user preferences; (2) User profiles cannot be updated dynamically: Traditional dialogue systems often have fixed user profiles, while user status often changes dynamically. Existing methods lack the ability to adjust user profiles according to user instructions. Summary of the Invention

[0004] In order to solve the problem of low personalization of large language models in the existing technology, the present invention provides a large language model personalized question and answer method based on retrieval enhancement generation. By parsing personal situation descriptions and specific questions, and performing retrieval enhancement through a user portrait knowledge retrieval enhancement database, personalized adaptation of large language model questions and answers can be achieved.

[0005] According to one aspect of the present invention, a personalized question answering method based on a large language model generated through retrieval enhancement is provided, comprising: Receive the client's conversation request and parse it into personal situation description and specific questions; Sending the specific question to a pre-built user portrait knowledge retrieval enhancement database to obtain retrieval enhancement data containing the user portrait, and merging the data with the personal situation description into contextual retrieval enhancement data containing the user portrait; Based on the contextual retrieval enhanced data including the user portrait and the specific question, a question instruction is generated and input into the large language model to obtain corresponding answer feedback data.

[0006] As a further technical solution, the pre-construction of the user portrait knowledge retrieval enhancement database includes: Through a series of dialogue-based question-answering processes, users’ personal information is collected and a user portrait survey file is generated; Use pre-trained word vector models to parse and extract features from user portrait survey files; Using a pre-trained large language model, the unstructured text information after parsing and feature extraction is converted into a structured "attribute-situation" user portrait data format; Save user portraits to the knowledge retrieval enhanced database.

[0007] As a further technical solution, the personal situation description and analysis of specific issues include: The client's conversation request is parsed based on a trained natural language processing deep learning model to obtain a personal description and specific questions. The personal description includes background information proposed in the conversation request to assist in improving the ability to provide personalized answers, and the specific questions include actual questions or requests proposed in the conversation request that the user hopes to have answered.

[0008] As a further technical solution, the specific question is sent to a pre-built user profile knowledge retrieval enhancement database to obtain retrieval enhancement data containing the user profile, and is combined with the personal situation description to form contextual retrieval enhancement data containing the user profile, including: Determining a question retrieval vector according to the specific question; Sending the question retrieval vector to a pre-built user portrait knowledge enhancement database to query the corresponding user portrait enhancement retrieval vector; The enhanced retrieval vector and the personal situation are combined into contextual retrieval enhanced data containing a user portrait.

[0009] As a further technical solution, querying the corresponding user portrait to enhance the retrieval vector includes: Calculate the similarity between each database vector and the question retrieval vector, filter out results that are greater than or equal to the similarity threshold, and remove duplicates to obtain an enhanced retrieval vector containing the user portrait.

[0010] As a further technical solution, based on the contextual retrieval enhanced data including the user profile and the specific question, a question instruction is generated and input into the large language model, including: Constructing question-answer prompt data based on the contextual retrieval enhanced data including the user portrait and the specific question; The question-and-answer prompt data is input into the large language model to obtain answer feedback data corresponding to the specific question.

[0011] As a further technical solution, question-answer prompt data is constructed based on the contextual retrieval enhanced data including the user portrait and the specific question, including: Get the preset {user profile, specific question} large model Q&A prompt template; The context retrieval enhancement data including the user portrait and the specific question are filled into the question and answer prompt template to generate personalized question and answer prompt data.

[0012] As a further technical solution, after obtaining the corresponding answer feedback data, it also includes: The user portrait knowledge retrieval enhancement database is updated based on the personal situation description.

[0013] As a further technical solution, the method further includes: Analyze the user's current emotional tendency based on the personal situation description; The current emotional tendency and timestamp are written into the user portrait retrieval enhancement database for subsequent calls.

[0014] According to one aspect of the present invention, a personalized question-answering system based on a large language model generated through retrieval enhancement is provided, comprising: The request parsing module is used to receive the client's conversation request and parse it into personal situation description and specific questions; A retrieval enhancement module is used to send the specific question to a pre-built user profile knowledge retrieval enhancement database, obtain retrieval enhancement data containing the user profile, and merge it with the personal situation description into contextual retrieval enhancement data containing the user profile; The personalized question-answering module is used to retrieve the enhanced data and the specific question based on the context containing the user portrait, generate question instructions and input them into the large language model to obtain corresponding answer feedback data.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. After receiving a conversation request from the client, the present invention parses the personal description and specific questions, and then performs retrieval enhancement in combination with the user portrait knowledge retrieval enhancement database. This can achieve personalized adaptation of large language model questions and answers. It has the advantages of strong adaptability, dynamic updating, and individual preference alignment. It can also be widely used in other large-model-based intelligent systems and has a good prospect for promotion and application.

[0016] 2. On the one hand, the present invention uses a database to store user portraits and obtains knowledge-enhanced context data through retrieval enhancement generation technology, which can ensure that user preferences are embedded in the large model dialogue system, ensure the integrity and diversity of the large language model input data, and thus enhance the ability of the large language model to generate personalized user preferences; on the other hand, after completing a complete conversation, the present invention will summarize the user's current emotions, recent events and other attributes, and write them into the database, thereby dynamically updating the user portrait in the database and forming a complete and comprehensive dynamic update of user preference generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction will be given below to the drawings used in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A flowchart of a personalized question-answering method using a large language model based on retrieval enhancement generation provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention are arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the sequence of steps and / or structural composition mode, but must be based on the ability of ordinary technicians in this field to implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that this combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0020] See Figure 1As shown, a schematic diagram of a personalized question-answering method based on a large language model generated based on retrieval enhancement is provided for an embodiment of this specification. In an embodiment of this specification, a personalized question-answering method based on a large language model generated based on retrieval enhancement can be applied to a third-party joint retrieval terminal. The third-party joint retrieval terminal or its operating environment is independent of the client and server providing the service. Specifically, the third-party joint retrieval terminal can be a terminal device or a server. This example embodiment does not make special restrictions on this. The following takes the third-party joint retrieval terminal as an example. Figure 1 The process shown in FIG. 1 is described in detail. The personalized question answering method based on the large language model generated by retrieval enhancement in the embodiment of this specification may specifically include the following steps: Step S1, pre-building a user portrait knowledge retrieval enhancement database; Step S2, receiving a conversation request from the client and parsing it into a personal description and specific questions; Step S3: sending the specific question to a knowledge retrieval enhancement database, obtaining retrieval enhancement data including the user portrait, and merging the data with the personal situation description into contextual retrieval enhancement data including the user portrait; Step S4: generating a question instruction based on the contextual retrieval enhanced data including the user portrait and the specific question and inputting the instruction into the large language model to obtain corresponding answer feedback data; Step S5: updating the user portrait knowledge retrieval enhancement database based on the personal situation description.

[0021] According to the large language model personalized question-answering method based on retrieval-enhanced generation in the embodiments of this specification, on the one hand, by using a database to store user portraits and obtaining knowledge-enhanced context data through retrieval-enhanced generation technology, it can be ensured that user preferences are embedded in the large model dialogue system, ensuring the integrity and diversity of the large language model input data, thereby improving the ability of the large language model to generate personalized user preferences; on the other hand, after completing a complete conversation, the system will summarize the user's current emotions, recent events and other attributes, and write them into the database, thereby dynamically updating the user portraits in the database and forming a complete and comprehensive dynamic update of user preference generation.

[0022] Below, the personalized question-answering method of a large language model based on retrieval enhancement generation in the embodiments of this specification is described in detail.

[0023] In step S1, a user portrait knowledge retrieval enhancement database is pre-built, and the user portrait knowledge retrieval enhancement database is a vector database.

[0024] In an example embodiment of the present specification, pre-constructing a user profile refers to user preference information collected through the dialogue system's question-and-answer process when the user first uses the system. For example, the initial dialogue may include a communicative understanding of the user's name, gender, age, and personality, such as when the large model asks "Are you a boy or a girl?"; question dialogues can also be obtained from inquiries, such as the user's instruction "I want you to play my lover." Of course, the dialogue system may also obtain preferences for other content, and this example embodiment does not specifically limit this.

[0025] In this embodiment, the conversational interaction method can be a software platform providing a user interface with an input box on the interface, where the user manually enters messages to interact. Alternatively, a voice recognition module can be provided to convert user voice information into text, thereby obtaining user input data. Alternatively, the user input command can be obtained by inputting images or files using visual multimodal technology. This embodiment does not specifically limit the current data input method. After obtaining the user input data, feedback can be generated based on the input data using a pre-deployed large model.

[0026] Specifically, when a user first uses this system, a series of dialogue-based question-and-answer processes are used to collect the user's personal information, form a comprehensive understanding of the user, and generate a user portrait survey file; a pre-trained word vector model is used to parse and extract features from the user portrait survey file; a pre-trained large language model is used to convert the unstructured text information after the parsing and feature extraction into a structured "attribute-situation" user portrait data format; and the user portrait is saved in a knowledge retrieval enhanced database.

[0027] The analysis and feature extraction process uses pre-trained models to deeply analyze user profiles, extracting various user attributes and their corresponding situations. For example, the user's age range, gender, specific gender preferences, personality type (e.g., extrovert, introvert), and relationship status (single, married, etc.) can be identified and categorized from their responses.

[0028] In step S2, a conversation request from the client is received and parsed into a personal situation description and specific questions.

[0029] In an example embodiment of the present specification, the conversation request is parsed using a natural language processing deep learning model deployed on the server side for performing text classification tasks. For example, the deep learning model can be a trained BERT (Bidirectional Encoder Representations from Transformers) model. Of course, the deep learning model can also be a trained CNN (Convolutional Neural Network) model. This example embodiment does not specifically limit the model used for parsing.

[0030] The personal profile refers to user-provided background information included in the client's conversation request. This profile is generally required to enhance the model's personalized question-and-answer capabilities. For example, it can include information a person is willing to provide, such as their recent mood, whether their work has been going well, or whether they've slept well. It can also include user-provided health information, such as allergies or chronic illnesses. It can also include user-provided lifestyle information, such as their eating habits or exercise frequency. This embodiment does not impose any specific restrictions on the content of the conversation request or the personal profile information it can provide.

[0031] A specific question refers to a practical question or request raised in a conversation request that the user hopes to have answered. It can cover a variety of areas. For example, a specific request can include medical consultation, such as "I've been having stomach pains lately. Which department should I go to?"; a specific request can also include emotional needs, such as "I'm having an argument with my parents. Please comfort me." Of course, a specific request can also be life advice, such as "I've been suffering from insomnia lately. How can I solve this problem?" This embodiment does not specifically limit the specific questions in the conversation request.

[0032] This step ensures accurate understanding and effective classification of user input by parsing the conversation request and breaking it down into the user's personal situation description and specific questions asked.

[0033] In step S3, the specific question is sent to a knowledge retrieval enhancement database to obtain retrieval enhancement data including the user portrait, and the retrieval enhancement data is merged with the personal situation description into contextual retrieval enhancement data including the user portrait.

[0034] In one embodiment of the present specification, a knowledge retrieval enhancement database is established in S1, which is a knowledge retrieval database containing user portraits. For example, the knowledge retrieval enhancement database can be a relational database, such as MySQL, Oracle database, etc., which stores user portrait information through databases, data tables, data rows and data fields; of course, the data retrieval enhancement database can also be a non-relational database, such as Mongodb database, etc., which stores user portraits in the form of tables. This embodiment does not specifically limit the specific type of the knowledge retrieval enhancement database.

[0035] Retrieval enhancement data is information related to the user profile retrieved from the knowledge retrieval enhancement database, which can help answer the user's specific questions. For example, when a user asks "What should I do if my parents quarrel?", the retrieval enhancement data retrieves relevant data vectors in the knowledge retrieval enhancement database, selects data vectors with high matching degrees, and determines the corresponding detection enhancement data based on these data vectors. For example, it may be possible to retrieve the record of the user's previous story about the quarrel between his parents, thereby obtaining more targeted and preferred content.

[0036] The specific question is sent to the knowledge enhancement database, and retrieval enhancement data including the user portrait is obtained, and the retrieval enhancement data including the personal situation description is merged into contextual retrieval enhancement data including the user portrait, including: determining a question retrieval vector according to the specific question; sending the question retrieval vector to the knowledge enhancement database, and querying the corresponding user portrait enhancement retrieval vector in the user portrait database; receiving the enhanced retrieval vector, and merging it with the personal situation into contextual retrieval data including the user portrait.

[0037] The database query process calculates the similarity between each database vector and the question retrieval vector, then filters the results that are greater than or equal to the similarity threshold, and finally removes duplicates from the filtered results.

[0038] Contextual search-enhanced data containing user profiles refers to data that is a combination of specific questions and search-enhanced data retrieved from the knowledge search-enhanced database. This data is then fed into a larger model to generate better answers. For example, this merging can be done by directly concatenating the data, or by labeling the search-enhanced data and specific questions separately. Of course, a larger language model can also be used to further integrate the two data sets. This embodiment does not impose any specific restrictions on the data merging process.

[0039] In step S4, based on the contextual retrieval enhancement data including the user portrait and the specific question, a question instruction is generated and input into the large language model to obtain corresponding answer feedback data.

[0040] In one embodiment of this specification, a large language model is a deep learning model pre-deployed on the server for performing various types of conversational tasks. For example, the large language model can be a trained unimodal GPT (Generative Pre-trained Transformer) model. Alternatively, the large language model can be a trained multimodal model combining text, vision, and speech. This embodiment does not impose any specific restrictions on the large language model's structure, training data, modality, or development company.

[0041] Based on the context retrieval enhanced data containing the user portrait and the specific question, a question instruction is generated and input into the large language model to obtain corresponding answer feedback data, including: constructing question and answer prompt data through the context retrieval enhanced data containing the user portrait and the specific question; inputting the question and answer prompt data into the large language model to obtain answer feedback data corresponding to the specific question.

[0042] The method of constructing question and answer prompt data through the context retrieval enhanced data containing the user portrait and the specific question includes: obtaining a preset {user portrait, specific question} large model question and answer prompt template; filling the context retrieval enhanced data containing the user portrait and the specific question into the question and answer prompt template to generate the personalized specific corresponding answer feedback data.

[0043] In step S5, the user portrait knowledge retrieval enhancement database is updated based on the personal situation description.

[0044] In one embodiment of this specification, after the overall answer is completed, the system updates the user profile's enhanced knowledge retrieval database based on the individual's description, thereby achieving personalized question-and-answer capabilities. Specifically, when a user shares a recent relevant event or makes a request to the system, if relevant content is not found in the enhanced knowledge retrieval database, the new information is written to the enhanced retrieval database, thereby optimizing the capabilities of the dialogue system during use. The update method can be to fix the specific information and add the newly added information; or it can be constructed in the format of {personal information, timestamp} and determine the freshness of the information based on time. This embodiment does not specifically limit the method for updating the user profile.

[0045] As an example, updating the user portrait knowledge retrieval enhancement database based on the personal situation description includes: analyzing the current emotion, status and other emotional tendencies according to the personal situation description obtained by parsing the dialogue request from the client; writing the current emotional tendencies and timestamps into the user portrait retrieval enhancement database for subsequent calls.

[0046] The implementation basis of each embodiment of the present invention is achieved through programmed processing by a device with processor functionality. Therefore, in practical engineering, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this reality, on the basis of the above-mentioned embodiments, an embodiment of the present invention provides a personalized question-answering system based on a large language model generated through retrieval enhancement. This system is used to implement the personalized question-answering method based on a large language model generated through retrieval enhancement described in the above-mentioned method embodiments.

[0047] The system includes: a request parsing module for receiving a dialogue request from a client and parsing it into a personal situation description and a specific question; a retrieval enhancement module for sending the specific question to a pre-built user portrait knowledge retrieval enhancement database, obtaining retrieval enhancement data containing the user portrait, and merging it with the personal situation description into contextual retrieval enhancement data containing the user portrait; a personalized question-and-answer module for generating question instructions based on the contextual retrieval enhancement data containing the user portrait and the specific question, and inputting them into a large language model to obtain corresponding answer feedback data.

[0048] The embodiment of the present invention provides a large language model personalized question-answering system based on retrieval enhancement generation. It addresses the problem that the large language model in the existing technology has a low degree of personalization. It adopts the aforementioned modules, analyzes personal situation descriptions and specific questions, and performs retrieval enhancement through a user portrait knowledge retrieval enhancement database, thereby realizing personalized adaptation of large language model questions and answers.

[0049] It should be noted that the system embodiments provided by the present invention are not only used to implement the methods in the above-mentioned method embodiments, but also used to implement the methods in other method embodiments provided by the present invention. The only difference lies in the setting of corresponding functional modules, and the principles thereof are basically the same as the principles of the above-mentioned system embodiments provided by the present invention. As long as those skilled in the art refer to the specific technical solutions in other method embodiments on the basis of the above-mentioned system embodiments, obtain corresponding technical means and technical solutions composed of these technical means by combining technical features, and on the premise of ensuring the practicality of the technical solutions, improve the modules in the above-mentioned system embodiments to obtain corresponding system class embodiments for implementing the methods in other method class embodiments.

[0050] In summary, the present invention discloses a method for personalized question-answering based on a large language model generated by retrieval enhancement. The method mainly includes the following steps: pre-building a user portrait knowledge retrieval enhancement database; accepting a dialogue request from a client, and parsing it into a personal situation description and a specific question; sending the specific question to the knowledge retrieval enhancement database, obtaining retrieval enhancement data containing the user portrait, and merging it with the personal situation description into contextual retrieval enhancement data containing the user portrait; based on the contextual retrieval enhancement data containing the user portrait and the specific question, generating a question instruction and inputting it into a large language model, obtaining corresponding answer feedback data, and updating the user portrait knowledge retrieval enhancement database based on the personal situation description. The present invention can realize personalized adaptation of large language model questions and answers, and has the advantages of strong adaptability, dynamic update, and personality preference alignment. It can also be widely used in other intelligent systems based on large models, and has a good prospect for promotion and application.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A personalized question-answering method based on a large language model generated by retrieval enhancement, characterized by: include: Receive the client's conversation request and parse it into personal situation description and specific questions; Sending the specific question to a pre-built user portrait knowledge retrieval enhancement database to obtain retrieval enhancement data containing the user portrait, and merging the data with the personal situation description into contextual retrieval enhancement data containing the user portrait; Based on the contextual retrieval enhanced data including the user portrait and the specific question, a question instruction is generated and input into the large language model to obtain corresponding answer feedback data.

2. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 1, characterized in that: The pre-construction of the user portrait knowledge retrieval enhancement database includes: Through a series of dialogue-based question-answering processes, users’ personal information is collected and a user portrait survey file is generated; Use pre-trained word vector models to parse and extract features from user portrait survey files; Using a pre-trained large language model, the unstructured text information after parsing and feature extraction is converted into a structured "attribute-situation" user portrait data format; Save user portraits to the knowledge retrieval enhanced database.

3. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 1 is characterized in that: A description of the individual's circumstances and analysis of specific issues, including: The client's conversation request is parsed based on a trained natural language processing deep learning model to obtain a personal description and specific questions. The personal description includes background information proposed in the conversation request to assist in improving the ability to provide personalized answers, and the specific questions include actual questions or requests proposed in the conversation request that the user hopes to have answered.

4. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 1, characterized in that: Send the specific question to a pre-built user profile knowledge retrieval enhancement database to obtain retrieval enhancement data containing the user profile, and merge it with the personal situation description into contextual retrieval enhancement data containing the user profile, including: Determining a question retrieval vector according to the specific question; Sending the question retrieval vector to a pre-built user portrait knowledge enhancement database to query the corresponding user portrait enhancement retrieval vector; The enhanced retrieval vector and the personal situation are combined into contextual retrieval enhanced data containing a user portrait.

5. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 4 is characterized in that: Query the corresponding user profile to enhance the search vector, including: Calculate the similarity between each database vector and the question retrieval vector, filter out results that are greater than or equal to the similarity threshold, and remove duplicates to obtain an enhanced retrieval vector containing the user portrait.

6. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 1, characterized in that: Generating a question instruction based on the contextual retrieval enhancement data including the user profile and the specific question and inputting it into the large language model includes: Constructing question-answer prompt data based on the contextual retrieval enhanced data including the user portrait and the specific question; The question-and-answer prompt data is input into the large language model to obtain answer feedback data corresponding to the specific question.

7. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 6, characterized in that: Constructing question and answer prompt data based on the contextual retrieval enhanced data including the user portrait and the specific question, including: Get the preset {user profile, specific question} large model Q&A prompt template; The context retrieval enhancement data including the user portrait and the specific question are filled into the question and answer prompt template to generate personalized question and answer prompt data.

8. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 1, characterized in that: After obtaining the corresponding answer feedback data, it also includes: The user portrait knowledge retrieval enhancement database is updated based on the personal situation description.

9. The personalized question-answering method based on a large language model generated by retrieval enhancement according to claim 8, characterized in that: The method further comprises: Analyze the user's current emotional tendency based on the personal situation description; The current emotional tendency and timestamp are written into the user portrait retrieval enhancement database for subsequent calls.

10. A personalized question-answering system based on a large language model generated by retrieval enhancement, characterized by: include: The request parsing module is used to receive the client's conversation request and parse it into personal situation description and specific questions; A retrieval enhancement module is used to send the specific question to a pre-built user profile knowledge retrieval enhancement database, obtain retrieval enhancement data containing the user profile, and merge it with the personal situation description into contextual retrieval enhancement data containing the user profile; The personalized question-answering module is used to retrieve the enhanced data and the specific question based on the context containing the user portrait, generate question instructions and input them into the large language model to obtain corresponding answer feedback data.

Images