User personalized knowledge base creating method, device, medium, equipment and system
By extracting multi-layered semantic features and user interaction behaviors from a large model knowledge base, a personalized knowledge base is constructed, which solves the problems of insufficient semantic association and user adaptability of general knowledge bases, and realizes efficient personalized knowledge services and real-time updates.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 凌川峰(贵州)信息技术有限公司
- Filing Date
- 2025-12-19
- Publication Date
- 2026-04-21
AI Technical Summary
In existing technologies, general knowledge bases struggle to effectively capture semantic relationships between knowledge points, lack adaptability to individual user differences, have insufficient retrieval accuracy and recall, and suffer from low efficiency in real-time incremental updates of large-scale knowledge bases.
By extracting multi-layered semantic features of text from a large model knowledge base, identifying entities and entity relationships, mapping them to a vector space, recording user interaction behavior, dynamically adjusting vector representations, and using knowledge distillation technology to construct a user-personalized knowledge base for localization and real-time updates.
It enables personalized knowledge services based on individual user needs, improves retrieval accuracy and recall, and enhances the real-time update efficiency of the knowledge base and user satisfaction.
Smart Images
Figure CN121901430A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of knowledge graphs, and more particularly to a method, apparatus, medium, device, and system for creating a user-personalized knowledge base. Background Technology
[0002] In related technologies, general knowledge bases use traditional vectorization methods, employing techniques such as TF-IDF and Word2Vec to transform knowledge content into vector representations. This makes it difficult to effectively capture the semantic relationships between knowledge points, resulting in insufficient retrieval accuracy and recall. General knowledge bases are built using unified standards, and their internal knowledge structures are stored in the form of knowledge graphs or relational databases. This lacks the ability to adapt to individual user differences and cannot provide personalized knowledge services based on users' usage habits and knowledge backgrounds. Furthermore, the real-time incremental updates of large-scale knowledge bases are inefficient and cannot meet the needs of dynamic application scenarios. Summary of the Invention
[0003] To overcome the problems existing in related technologies, this application provides a method, apparatus, medium, device and system for creating a user-personalized knowledge base.
[0004] According to a first aspect of this application, a method for creating a user-personalized knowledge base is provided, applicable to a large model knowledge base, including: Collect knowledge input containing text, extract multi-layer semantic features of the text, identify entities and entity relationships in the text, and map the entities and entity relationships to a vector space; capture complex relationships between entities, and fuse them with the multi-layer semantic features of the text to form vector representations of various entities and entity relationships; Record user interaction behavior with the large model knowledge base, extract user preference features, and classify user knowledge interests into corresponding levels and domains based on the user preference features; Based on the level and domain of the user's knowledge interests, relevant vector representations are extracted from the large model knowledge base to construct a personalized knowledge base for the user.
[0005] Based on the foregoing solution, in some embodiments of this application, the method for creating a user-personalized knowledge base further includes: The vector representation is dynamically adjusted based on different queries and usage scenarios, so that the same entity has different vector representations in different scenarios.
[0006] Based on the foregoing scheme, in some embodiments of this application, the step of extracting relevant vector representations from the large model knowledge base according to the level and domain of the user's knowledge interests to construct a personalized user knowledge base includes: Using knowledge distillation technology, vector representations related to the level and domain of the user's knowledge interests are extracted from a large model knowledge base to construct a personalized knowledge base for the user.
[0007] Based on the foregoing solution, in some embodiments of this application, the method for creating a user-personalized knowledge base further includes: Acquire new knowledge and transform it into a new vector; The new vector is inserted into the vector space, connected to the nearest neighbor entity, and the old vector is marked as invalid, thereby updating the large model knowledge base; The new vector is compared with the level and domain to which the user's knowledge interests belong to determine the relevance of the new vector; when the relevance is greater than or equal to a preset threshold, the new vector is used to update the user's personalized knowledge base.
[0008] Based on the foregoing solution, in some embodiments of this application, the method for creating a user-personalized knowledge base further includes: Based on the user's personalized knowledge base, the user's knowledge items are determined. The knowledge items include knowledge points already mastered, areas or topics of interest, specific entities or concepts, user preferences or attributes. Develop a personalized forgetting model for each knowledge item; Monitor the user's retention time for different knowledge items and dynamically adjust the parameters of the forgetting model.
[0009] Based on the aforementioned solution, in some embodiments of this application, the user-personalized knowledge base is deployed on the client side to perform localized processing of user-sensitive data.
[0010] According to another aspect of this application, a user-personalized knowledge base creation apparatus is provided, comprising: The vector database service module is used to collect knowledge input containing text, extract multi-layer semantic features of the text, identify entities and entity relationships in the text, and map the entities and entity relationships to a vector space; capture complex relationships between entities, and fuse them with the multi-layer semantic features of the text to form vector representations of various entities and entity relationships; The user profile generation module is used to record the user's interaction behavior with the large model knowledge base, extract user preference features, and divide the user's knowledge interests into corresponding levels and domains based on the user preference features. The personalized knowledge base generation module is used to extract relevant vector representations from the large model knowledge base according to the level and domain of the user's knowledge interests, and construct the user's personalized knowledge base.
[0011] Based on the foregoing solution, in some embodiments of this application, the vector database service module is further configured to: The vector representation is dynamically adjusted based on different queries and usage scenarios, so that the same entity has different vector representations in different scenarios.
[0012] Based on the foregoing solution, in some embodiments of this application, the vector database service module is further configured to: Acquire new knowledge and transform it into a new vector; The new vector is inserted into the vector space, connected to the nearest neighbor entity, and the old vector is marked as invalid, thereby updating the large model knowledge base; The new vector is compared with the level and domain to which the user's knowledge interests belong to determine the relevance of the new vector; when the relevance is greater than or equal to a preset threshold, the new vector is used to update the user's personalized knowledge base.
[0013] Based on the foregoing solution, in some embodiments of this application, the personalized knowledge base generation module is further used for: Based on the user's personalized knowledge base, the user's knowledge items are determined. The knowledge items include knowledge points already mastered, areas or topics of interest, specific entities or concepts, user preferences or attributes. Develop a personalized forgetting model for each knowledge item; Monitor the user's retention time for different knowledge items and dynamically adjust the parameters of the forgetting model.
[0014] Based on the foregoing solution, in some embodiments of this application, the personalized knowledge base generation module is further used for: The user-personalized knowledge base is deployed on the client side to perform localized processing of sensitive user data.
[0015] According to another aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed, implements the steps of a method for creating a user-personalized knowledge base.
[0016] According to another aspect of this application, a computer device is provided, including a processor, a memory, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of a user-personalized knowledge base creation method.
[0017] According to another aspect of this application, a user-personalized knowledge base creation system is provided, comprising: a server and a client, wherein the server and the client are connected via a network; the client is configured to send knowledge input and interaction behavior data to the server and receive a user-personalized knowledge base constructed by deploying the server; the server is configured to run a large model knowledge base, execute a user-personalized knowledge base creation method to construct the user-personalized knowledge base, and send the user-personalized knowledge base to the client.
[0018] The client includes a browser running on a terminal device; the browser is used to load web pages to provide an interactive interface, collect the knowledge input through the interactive interface, and send the knowledge input to the server via HTTP / HTTPS protocol; and the user-personalized knowledge base is configured to be loaded into the browser's local cache or runtime environment.
[0019] This application extracts multi-layered semantic features from user-input text using a large-scale model knowledge base, identifies entities and entity relationships within the text, maps them to a vector space, and fuses them with the multi-layered semantic features to form vector representations of various entities and entity relationships. Furthermore, by recording user interactions with the large-scale model knowledge base, it extracts user preference features, categorizes user knowledge interests into corresponding levels and domains, and performs knowledge distillation on the large-scale model knowledge base based on these levels and domains. This distillation transforms the large-scale model knowledge base into a small, efficient student model, effectively transferring knowledge learned in the large-scale model's complex parameters to the small student model. This effectively captures semantic relationships between knowledge points, improving retrieval accuracy and recall. It also enables the creation of personalized knowledge bases based on individual user needs, providing personalized knowledge services according to user habits and knowledge background. Additionally, the small model can be deployed on the client side for localized processing of sensitive user data and the establishment of personalized forgetting models for each knowledge item, improving knowledge utilization.
[0020] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description
[0021] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating a method for creating a user-personalized knowledge base according to an exemplary embodiment.
[0022] Figure 2 This is a schematic diagram illustrating a user-personalized knowledge base creation apparatus according to an exemplary embodiment.
[0023] Figure 3 This is a block diagram illustrating a computer device for creating a user-personalized knowledge base according to an exemplary embodiment. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily combined with each other.
[0025] General-purpose knowledge bases employ traditional vectorization methods, using techniques such as TF-IDF (a common weighting technique in information retrieval and data mining used to evaluate the importance of a word to a document in a document set or corpus; TF stands for Term Frequency, and IDF stands for Inverse Document Frequency) and Word2Vec (a relevant model used to generate word vectors) to transform knowledge content into vector representations. This approach struggles to effectively capture semantic relationships between knowledge points, resulting in insufficient retrieval precision and recall. Furthermore, general-purpose knowledge bases are built using uniform standards, with internal knowledge structures stored in the form of knowledge graphs or relational databases. This lack of adaptability to individual user differences prevents the provision of personalized knowledge services based on user habits and knowledge backgrounds. Real-time incremental updates of large-scale knowledge bases are inefficient and fail to meet the demands of dynamic application scenarios.
[0026] To address the problems existing in the prior art, this application provides a method for creating a user-personalized knowledge base.
[0027] Figure 1 This is a flowchart illustrating a method for creating a user-personalized knowledge base according to an exemplary embodiment. (Reference) Figure 1 The methods for creating a user-personalized knowledge base include at least the following: Step S11: The large model knowledge base collects knowledge input containing text, extracts multi-layer semantic features of the text, identifies entities and entity relationships in the text, and maps the entities and entity relationships to a vector space; it captures complex relationships between entities and fuses them with the multi-layer semantic features of the text to form vector representations of various entities and entity relationships.
[0028] By setting up a large language model (hereinafter referred to as the large model) and a large model knowledge base on the server, various user queries can be received. For example, users can obtain relevant knowledge content through natural language questions or keyword searches.
[0029] The large model knowledge base collects knowledge inputs containing text. These inputs can be pre-entered knowledge content, user-inputted content, or relevant content retrieved from the internet based on user input.
[0030] Extract multi-layered semantic features from text to identify entities and entity relationships.
[0031] Multi-level semantic features include: word-level semantic features, sentence-level semantic features, paragraph-level semantic features, and document-level semantic features.
[0032] Word-level semantic features: Focus on the smallest meaningful unit of text, namely the attributes of a single word or tag.
[0033] Sentence-level semantic features: Focus on the arrangement of words within a sentence, grammatical structure, and semantic roles, processing the complete ideas of a single sentence.
[0034] Paragraph-level semantic features: Focus on the contextual relationships between sentences, resolve cross-sentence references and ambiguities, and process information within a thematic unit.
[0035] Document-level semantic features: Focus on the macro structure, theme, and global context of the entire document, and process the overall knowledge conveyed by the document.
[0036] Entities are the most basic nodes in a knowledge graph, representing distinguishable and independently existing things in the real world. An entity can be a word or a series of words, which always refer to a certain thing.
[0037] Entity relations fundamentally describe the semantic connections, associations, or interactions that exist between two or more entities identified in a given text.
[0038] In the standard representation of knowledge graphs, each fact is represented in the form of (entity 1, relation, entity 2), also known as a triple, forming an edge (relation) that connects entity nodes (entity) in the graph structure.
[0039] Mapping the identified entities and entity relationships to a vector space allows graph neural networks to learn graph structure features, capturing complex relationships between entities. This, combined with multi-layered semantic features of the text, generates various vector representations of entities and entity relationships. By extracting word-level, sentence-level, paragraph-level, and document-level semantic features layer by layer, each higher-level feature adds richer and deeper semantic information to entities and entity relationships. Therefore, each entity or entity relationship can include multiple semantic information and can have multiple vector representations.
[0040] Step S12: Record the user's interaction behavior with the large model knowledge base, extract user preference features, and divide the user's knowledge interests into corresponding levels and domains based on the user preference features.
[0041] By recording user queries to a large model or their keyword searches and other interactive behaviors, we can determine each user's preference characteristics, the areas of interest they have, and their level within those areas. For example, if user A frequently searches for mathematical questions involving only general algebra or geometry, then user A's knowledge interest can be identified as mathematics at a beginner level. If user B frequently searches for calculus-related questions, then user B's knowledge interest can be identified as mathematics at an advanced level.
[0042] Step S13: Based on the level and domain of the user's knowledge interests, extract relevant vector representations from the large model knowledge base to construct a personalized knowledge base for the user.
[0043] Users have different areas of interest and require different knowledge. For example, if a user is interested in mathematics, then knowledge in history and humanities may not be what they need. Furthermore, users have different levels of knowledge, meaning the same knowledge entity can have different semantic representations, i.e., different vector representations. Therefore, based on the user's level and domain of knowledge interest, relevant vector representations are extracted from a large model knowledge base to create a personalized knowledge base for each user. Thus, by accessing this personalized knowledge base, users can efficiently and accurately obtain content related to their own level of knowledge and knowledge domain, providing personalized services.
[0044] In one exemplary embodiment, the method for creating a user-personalized knowledge base further includes: The vector representation is dynamically adjusted based on different queries and usage scenarios, so that the same entity has different vector representations in different scenarios.
[0045] In reality, the same entity can represent different things or have different meanings in different fields or scenarios. For example, "apple" can represent an electronic product or a company name in the technology field; in other fields, it represents a specific fruit. Therefore, in a large model knowledge base, the entity "apple" can have multiple entity relationships with other entities and multiple vector representations. When a user queries the large model knowledge base, the vector representation can be dynamically adjusted according to different queries and usage scenarios, so that the same entity has different vector representations in different scenarios. Then, based on the specific scenario of the user's query, a vector representation relevant to the query scenario is provided to the user. This can effectively improve retrieval accuracy and recall, and enhance user satisfaction.
[0046] In one exemplary embodiment, constructing a personalized knowledge base for a user by extracting relevant vector representations from a large model knowledge base based on the level and domain of the user's knowledge interests includes: By employing knowledge distillation technology, vector representations related to the level and domain of the user's knowledge interests are extracted from a large model knowledge base to construct a personalized knowledge base for the user.
[0047] When users perform queries and searches on the large model knowledge base, their interaction behavior with the large model knowledge base can be recorded, including queries, clicks, dwell time, and favorites. This allows for the analysis of users' query patterns and language habits, extraction of user preference features, determination of the level and domain of users' knowledge interests, identification of each query and usage scenario, and dynamic adjustment of vector representation.
[0048] By introducing knowledge distillation technology, vector representations related to the user's knowledge interests in terms of level and domain are extracted from a large model knowledge base and compressed into lightweight vector representations to build a personalized knowledge base for the user. Knowledge distillation can transfer the "knowledge" already learned by the large model and implicit in its complex parameters to a smaller model, enabling the utilization of the rich knowledge contained in the large model at a lower cost and with higher efficiency.
[0049] In one exemplary embodiment, the method for creating a user-personalized knowledge base further includes: The large model knowledge base acquires new knowledge and transforms the new knowledge into new vectors.
[0050] The new vector is inserted into the vector space, connected to the nearest neighbor entity, and the old vector is marked as invalid, thus updating the knowledge base of the large model.
[0051] The new vector is compared with the level and domain to which the user's knowledge interests belong to determine the relevance of the new vector; when the relevance is greater than or equal to a preset threshold, the new vector is used to update the user's personalized knowledge base.
[0052] When there are knowledge updates or knowledge conflicts, the large model knowledge base receives new knowledge. For example, when a new event occurs in a certain field, or a new record is generated, the existing knowledge needs to be updated to reflect the new information. The large model database can proactively acquire new knowledge and update existing knowledge; for example, it can continuously acquire the latest knowledge and proactively update the knowledge within its own database. Alternatively, if a user finds a conflict between the knowledge content they obtained from a personalized knowledge base and the latest knowledge, the user can report the conflict to the large model knowledge base. For example, if a user obtains a world record for a sport from a personalized knowledge base and discovers that the record is not the latest, the user can report the conflict by either submitting a "report error" message or directly inputting the corrected world record. The large model knowledge base then evaluates the accuracy and authority of the user's feedback to ensure it is not malicious or a false alarm before acquiring the new world record.
[0053] The large model knowledge base transforms new knowledge into new vectors. These new vectors are then inserted into the vector space and linked to their nearest neighbors. Old vectors are marked as invalid, thus updating the large model database.
[0054] After the large model database is updated, the new vectors are compared with the levels and domains to which the user's knowledge interests belong to determine the relevance of the new vectors; when the relevance is greater than or equal to a preset threshold, the new vectors are used to update the user's personalized knowledge base.
[0055] When new knowledge is generated, or when old knowledge conflicts with new knowledge, the large model knowledge base updates the knowledge. After the large model knowledge base updates the knowledge, the personalized knowledge base of the corresponding user is updated accordingly.
[0056] In one exemplary embodiment, a user-personalized knowledge base is deployed on the client side to perform localized processing of user-sensitive data.
[0057] After creating a personalized knowledge base for each user, this knowledge base can be deployed on the client-side. When the client performs knowledge queries through a webpage, it can access the locally deployed personalized knowledge base using pre-defined models within the webpage. This allows for the rapid and accurate acquisition of vector representations related to the user's knowledge interests and relevant domains, providing precise query results. Simultaneously, data processing is completed locally, significantly enhancing user privacy protection, reducing network latency, and minimizing server-side computation and transmission costs.
[0058] In one exemplary embodiment, when a small model accesses a locally deployed personalized knowledge base, and the personalized knowledge base does not contain the knowledge content queried by the user, it can query the large model knowledge base to obtain the relevant content. Simultaneously, the large model knowledge base determines the query and usage scenario based on the level and domain of the user's query content, dynamically adjusts the vector representation, and then employs knowledge distillation technology again to distill the adjusted dynamic vector representation into the personalized knowledge base, updating and expanding the personalized knowledge base to keep it synchronized with the user's learning content.
[0059] In one exemplary embodiment, the method for creating a user-personalized knowledge base further includes: Based on the user's personalized knowledge base, the user's knowledge items are determined. These knowledge items include knowledge points already mastered, areas or topics of interest, specific entities or concepts, and user preferences or attributes.
[0060] Develop a personalized forgetting model for each knowledge item.
[0061] Monitor the retention time of users' memories of different knowledge items and dynamically adjust the parameters of the forgetting model.
[0062] A user-personalized knowledge base is a collection of knowledge extracted from a large model knowledge base using knowledge distillation techniques, based on factors such as the user's browsing history, the level and domain of the user's knowledge interests. The knowledge items contained in the user-personalized knowledge base are the user's knowledge items, which include mastered knowledge points, areas or topics of interest, specific entities or concepts, and user preferences or attributes.
[0063] When a user first encounters a piece of knowledge, an initial memory trace is formed. Due to differences in the difficulty of the knowledge itself, its significance to the user, and the user's learning methods, the strength of this memory trace (memory strength) varies. Memory strength decreases over time. To enable users to effectively master various types of knowledge, a personalized forgetting model is established for each knowledge item. For example, an Ebbinghaus forgetting curve can be established. By monitoring user interaction data, the memory strength of each knowledge item can be dynamically estimated, predicting when the memory retention rate will drop below a certain threshold—the point at which forgetting is imminent. By monitoring the user's memory retention time for different knowledge items, dynamically adjusting the parameters of the forgetting model, and designing differentiated repetitive reminder mechanisms, users can better memorize various types of knowledge.
[0064] The method for creating a personalized knowledge base provided in this application will be further described below with specific embodiments. Specific Implementation Example 1 The large model database collects knowledge from various fields and user input containing text, extracts multi-layer semantic features from the text, identifies entities and entity relationships in the text, and maps entities and entity relationships to vector space; it captures complex relationships between entities, integrates them with multi-layer semantic features, and forms vector representations of various entities and entity relationships involving multiple fields.
[0066] It receives user access and queries, records user interaction behavior with a large model knowledge base, extracts user preference features, creates a profile for each user, and divides user knowledge interests into corresponding levels and domains.
[0067] Based on the level and domain of the user's knowledge interests, the query and usage scenarios are determined, and the vector representation is dynamically adjusted. Using knowledge distillation technology, the adjusted vector representations related to the level and domain of the user's knowledge interests are extracted from the large model knowledge base to construct a personalized knowledge base for the user.
[0068] Deploy a personalized knowledge base to the client.
[0069] Users can access knowledge queries by visiting web pages and use pre-set mini-models on the web pages to access locally deployed personalized knowledge bases.
[0070] When users find that the query results are inconsistent with the latest factual results, they can report the error to the large model database through the "Report Error" function.
[0071] The large model knowledge base evaluates the accuracy and authority of user feedback to ensure it is not malicious before obtaining the latest factual results. These new factual results are transformed into new vectors, inserted into the vector space, connected to the nearest neighbor entities, and old vectors are marked as invalid. Without modifying the core parameters of the large model, a new knowledge graph is dynamically constructed to store the revised correct knowledge and resolve knowledge conflicts.
[0072] After the large model knowledge base is updated, the new vectors are compared with the levels and domains to which the user's knowledge interests belong to determine the relevance of the new vectors. When the relevance is greater than or equal to a preset threshold, the new vectors are used to update the user's personalized knowledge base. In this process, not only the personalized knowledge base of the user who reported "error" is updated, but also the personalized knowledge bases of other users who contain the updated knowledge are updated.
[0073] Based on the user's personalized knowledge base, the system identifies the user's knowledge items, including mastered knowledge points, areas or topics of interest, specific entities or concepts, and user preferences or attributes. A personalized forgetting model is established for each knowledge item. The system monitors the user's memory retention time for different knowledge items and dynamically adjusts the parameters of the forgetting model to enable users to better remember various knowledge and improve the utilization rate of knowledge.
[0074] Through the above embodiments, the user-personalized knowledge base creation method provided in this application extracts multi-level semantic features of text, maps the identified entities and entity relationships to a vector space, and fuses the complex associations between entities and multi-level semantic features to effectively capture the semantic relationships between knowledge. It dynamically adjusts the vector representation according to different query and usage scenarios, so that the same entity has different vector representations in different scenarios, thereby improving retrieval accuracy and recall. It can also provide personalized knowledge services based on users' usage habits and knowledge background, and can efficiently update knowledge in real time.
[0075] Figure 2 This is a schematic diagram illustrating a user-personalized knowledge base creation apparatus according to an exemplary embodiment. (Reference) Figure 2 The user-personalized knowledge base creation device includes: a vector database service module 201, a user profile generation module 202, and a personalized knowledge base generation module 203.
[0076] The vector database service module 201 is configured to collect knowledge input containing text, extract multi-layer semantic features of the text, identify entities and entity relationships in the text, map entities and entity relationships to vector space, capture complex relationships between entities, and fuse them with the multi-layer semantic features of the text to form vector representations of various entities and entity relationships.
[0077] The user profile generation module 202 is configured to record the user's interaction with the large model knowledge base, extract user preference features, and divide the user's knowledge interests into corresponding levels and domains based on the user preference features.
[0078] The personalized knowledge base generation module 203 is configured to extract relevant vector representations from the large model knowledge base based on the level and domain of the user's knowledge interests, and construct a personalized knowledge base for the user.
[0079] In one exemplary embodiment, the vector database service module 201 is further configured to: This is used to dynamically adjust the vector representation based on different queries and usage scenarios, so that the same entity has different vector representations in different scenarios.
[0080] In one exemplary embodiment, the vector database service module 201 is further configured to: It is used to acquire new knowledge and transform new knowledge into new vectors.
[0081] The new vector is inserted into the vector space, connected to the nearest neighbor entity, and the old vector is marked as invalid, thereby updating the large model knowledge base; The new vector is compared with the level and domain to which the user's knowledge interests belong to determine the relevance of the new vector; when the relevance is greater than or equal to a preset threshold, the new vector is used to update the user's personalized knowledge base.
[0082] In one exemplary embodiment, the personalized knowledge base generation module 203 is further configured to: This is used to determine a user's knowledge items based on the user's personalized knowledge base. The knowledge items include knowledge points already mastered, areas or topics of interest, specific entities or concepts, and user preferences or attributes.
[0083] Develop a personalized forgetting model for each knowledge item.
[0084] Monitor the user's retention time for different knowledge items and dynamically adjust the parameters of the forgetting model.
[0085] In one exemplary embodiment, the personalized knowledge base generation module 203 is further configured to: This is used to deploy a personalized knowledge base on the client side and to perform localized processing of sensitive user data.
[0086] Figure 3 This is a block diagram illustrating a computer device 300 for creating a user-personalized knowledge base, according to an exemplary embodiment. For example, the computer device 300 may be provided as a server. (Refer to...) Figure 3 The computer device 300 includes a processor 301, the number of which can be set to one or more as needed. The computer device 300 also includes a memory 302 for storing instructions executable by the processor 301, such as application programs. The number of memories can be set to one or more as needed. The stored application programs can be one or more. The processor 301 is configured to execute instructions to perform the aforementioned user-personalized knowledge base creation method.
[0087] This application also provides a user-personalized knowledge base creation system, including: a server and a client, the server and the client being connected via a network; the client being used to send knowledge input and interaction behavior data to the server, and to receive the user-personalized knowledge base constructed by the server; the server being used to run a large model knowledge base, execute the method for creating the user-personalized knowledge base, to construct the user-personalized knowledge base, and send the user-personalized knowledge base to the client.
[0088] The client includes a browser running on a terminal device; the browser is used to load web pages to provide an interactive interface, collect knowledge input through the interactive interface, and send the knowledge input to the server via HTTP / HTTPS protocol; and the user's personalized knowledge base is configured to be loaded into the browser's local cache or runtime environment.
[0089] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus (devices), or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Computer storage media include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data), including but not limited to RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and can include any information delivery medium.
[0090] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0091] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0092] In this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the article or device that includes said element.
[0093] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.
[0094] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations fall within the scope of the claims of this application and their equivalents, the intent of this application also includes these modifications and variations.
Claims
1. A method for creating a user-personalized knowledge base, characterized in that, Applied to large model knowledge bases, including: Collect knowledge input containing text, extract multi-layer semantic features of the text, identify entities and entity relationships in the text, and map the entities and entity relationships to a vector space; capture complex relationships between entities, and fuse them with the multi-layer semantic features of the text to form vector representations of various entities and entity relationships; Record user interaction behavior with the large model knowledge base, extract user preference features, and classify user knowledge interests into corresponding levels and domains based on the user preference features; Based on the level and domain of the user's knowledge interests, relevant vector representations are extracted from the large model knowledge base to construct a personalized knowledge base for the user.
2. The method for creating a user-personalized knowledge base as described in claim 1, characterized in that, Also includes: The vector representation is dynamically adjusted based on different queries and usage scenarios, so that the same entity has different vector representations in different scenarios.
3. The method for creating a user-personalized knowledge base as described in claim 1, characterized in that, The step of extracting relevant vector representations from the large model knowledge base based on the level and domain of the user's knowledge interests to construct a personalized knowledge base includes: Using knowledge distillation technology, vector representations related to the level and domain of the user's knowledge interests are extracted from a large model knowledge base to construct a personalized knowledge base for the user.
4. The method for creating a user-personalized knowledge base as described in claim 1, characterized in that, Also includes: Acquire new knowledge and transform it into a new vector; The new vector is inserted into the vector space, connected to the nearest neighbor entity, and the old vector is marked as invalid, thereby updating the large model knowledge base; The relevance of the new vector is determined by comparing it with the level and domain to which the user's knowledge interests belong. When the correlation is greater than or equal to a preset threshold, the user-personalized knowledge base is updated using the new vector.
5. The method for creating a user-personalized knowledge base as described in claim 1, characterized in that, Also includes: Based on the user's personalized knowledge base, the user's knowledge items are determined. The knowledge items include knowledge points already mastered, areas or topics of interest, specific entities or concepts, user preferences or attributes. Develop a personalized forgetting model for each knowledge item; Monitor the user's retention time for different knowledge items and dynamically adjust the parameters of the forgetting model.
6. The method for creating a user-personalized knowledge base as described in claim 1, characterized in that, The user-personalized knowledge base is deployed on the client side to perform localized processing of sensitive user data.
7. A user-personalized knowledge base creation device, characterized in that, include: The vector database service module is used to collect knowledge input containing text, extract multi-layer semantic features of the text, identify entities and entity relationships in the text, and map the entities and entity relationships to a vector space; capture complex relationships between entities, and fuse them with the multi-layer semantic features of the text to form vector representations of various entities and entity relationships; The user profile generation module is used to record the user's interaction behavior with the large model knowledge base, extract user preference features, and divide the user's knowledge interests into corresponding levels and domains based on the user preference features. The personalized knowledge base generation module is used to extract relevant vector representations from the large model knowledge base according to the level and domain of the user's knowledge interests, and construct the user's personalized knowledge base.
8. The user-personalized knowledge base creation device as described in claim 7, characterized in that, The vector database service module is also used for: The vector representation is dynamically adjusted based on different queries and usage scenarios, so that the same entity has different vector representations in different scenarios.
9. The user-personalized knowledge base creation device as described in claim 7, characterized in that, The vector database service module is also used for: Acquire new knowledge and transform it into a new vector; The new vector is inserted into the vector space, connected to the nearest neighbor entity, and the old vector is marked as invalid, thereby updating the large model knowledge base; The relevance of the new vector is determined by comparing it with the level and domain to which the user's knowledge interests belong. When the correlation is greater than or equal to a preset threshold, the user-personalized knowledge base is updated using the new vector.
10. The user-personalized knowledge base creation device as described in claim 7, characterized in that, The personalized knowledge base generation module is also used for: Based on the user's personalized knowledge base, the user's knowledge items are determined. The knowledge items include knowledge points already mastered, areas or topics of interest, specific entities or concepts, user preferences or attributes. Develop a personalized forgetting model for each knowledge item; Monitor the user's retention time for different knowledge items and dynamically adjust the parameters of the forgetting model.
11. The user-personalized knowledge base creation device as described in claim 7, characterized in that, The personalized knowledge base generation module is also used for: The user-personalized knowledge base is deployed on the client side to perform localized processing of sensitive user data.
12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the method as described in any one of claims 1-6.
13. A computer device comprising a processor, a memory, and a computer program stored in the memory, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-6.
14. A user-personalized knowledge base creation system, characterized in that, include: A server and a client, wherein the server and the client are connected via a network; The client is used to send knowledge input and interaction behavior data to the server, and to receive the user-personalized knowledge base built by deploying the server. The server is configured to run a large model knowledge base, execute the method described in any one of claims 1 to 6 to construct the user-personalized knowledge base, and send the user-personalized knowledge base to the client.
15. The user-personalized knowledge base creation system as described in claim 14, characterized in that, The client includes a browser running on a terminal device; the browser is used to load web pages to provide an interactive interface, collect the knowledge input through the interactive interface, and send the knowledge input to the server via HTTP / HTTPS protocol; and the user-personalized knowledge base is configured to be loaded into the browser's local cache or runtime environment.