A large model-based medical personal knowledge base management method, system, device and medium

Abstract

The application discloses a medical personal knowledge base management method and system based on a large model, equipment and a medium, mainly related to the technical field of knowledge base management, to solve the problems that the existing scheme has multiple sources of medical knowledge, it is difficult to realize cross-modal unified modeling and integrated analysis, it lacks personalized knowledge organization and calling mechanism, and it cannot support context understanding and weak retrieval ability. Including: configuring a semantic question and answer framework in a reading interface, obtaining actual retrieval requirements through an input box in the semantic question and answer framework; obtaining semantic structure and keyword information by analyzing the actual retrieval requirements through a large model; determining corresponding knowledge segments from a preset knowledge database according to the relationship between the keyword information and the knowledge label vector; and obtaining an output answer result by using a medical AI large model based on the corresponding knowledge segments, the actual retrieval requirements and the preset knowledge database.

Description

Technical Field

[0001] This application relates to the field of knowledge base management technology, and in particular to a method, system, device and medium for managing a medical personal knowledge base based on a large model. Background Technology

[0002] In existing technologies, medical knowledge management systems are primarily geared towards institutional or general medical databases, lacking the ability to provide personalized management and intelligent support for the medical knowledge of individual users such as doctors and researchers. Specifically, the following technical problems exist:

[0003] 1. Medical knowledge comes from diverse sources and has a complex structure, making it difficult to manage in a unified manner: Medical knowledge involves multimodal and multi-source content such as medical records, examination reports, medical literature, textbooks, academic notes, and imaging data. Existing systems generally struggle to achieve unified modeling and integrated analysis across modalities.

[0004] 2. Lack of personalized knowledge organization and retrieval mechanisms: Currently, most knowledge bases adopt static and structured storage methods, which cannot dynamically adjust the knowledge content or presentation method according to the user's specialization, focus, learning path, etc., thus limiting personalized learning and decision support.

[0005] 3. Weak retrieval capabilities, unable to achieve semantic-level understanding and accurate question answering: Traditional keyword retrieval methods are difficult to meet the needs of highly complex knowledge queries in clinical or scientific research, and cannot support contextual understanding. Summary of the Invention

[0006] This application provides a method, system, device, and medium for managing a medical personal knowledge base based on a large model, in order to solve the problems of existing solutions, such as the difficulty in achieving cross-modal unified modeling and integrated analysis due to the diverse sources of medical knowledge, the lack of personalized knowledge organization and retrieval mechanisms, and the inability to support contextual understanding retrieval capabilities.

[0007] Firstly, this application provides a method for managing a medical personal knowledge base based on a large model, the method comprising:

[0008] Based on the current user's individual knowledge needs, determine the medical data acquisition interface and acquire medical data from several sources;

[0009] Based on the specific type of medical data, determine the corresponding pre-trained large model; using the pre-trained large model, extract knowledge fragments with pre-defined standardized descriptions from the medical data, and store the knowledge fragments and the corresponding URLs of the medical data in a pre-defined knowledge database; extract knowledge tags from the knowledge fragments;

[0010] Collect multi-dimensional exclusive tags for the current user; these tags should include at least: professional field, topics of interest, and behavioral preferences.

[0011] Vectorize the current user's knowledge tags and multi-dimensional exclusive tags to obtain knowledge tag and multi-dimensional exclusive tag vectors; add the knowledge tag vectors and the relationship between knowledge tag vectors and knowledge fragments to the preset knowledge database; calculate the similarity value between knowledge tag vectors and exclusive tag vectors;

[0012] Based on similarity values, determine the arrangement and recommendation order of knowledge fragments on the current user's reading interface;

[0013] Configure a semantic question-answering framework in the reading interface, and obtain the actual search requirements through the input boxes in the semantic question-answering framework;

[0014] By analyzing actual retrieval needs through large-scale models, semantic structure and keyword information are obtained;

[0015] Based on the relationship between keyword information and knowledge tag vectors, the corresponding knowledge fragments are determined from a pre-set knowledge database;

[0016] Based on the corresponding knowledge fragments, actual retrieval needs, and a pre-set knowledge database, a large-scale medical AI model is used to obtain the output answer results.

[0017] In one implementation of this application, based on the user's current personal knowledge needs, the retrieval information of the medical data acquisition interface is determined, and medical data from several sources is acquired, specifically including:

[0018] Input your personal knowledge needs into a standardized AI model to obtain search information;

[0019] Input the search information into various medical data acquisition interfaces to obtain medical data from several sources that meet your personal knowledge needs.

[0020] In one implementation of this application, a corresponding pre-trained large model is determined based on the specific type of medical data; using the pre-trained large model, knowledge fragments with pre-defined standardized descriptions are extracted from the medical data, specifically including:

[0021] When medical data is in text format, a pre-trained large medical model is used to extract entity data and the relationships between entity data from the text data.

[0022] Convert the descriptive terms involved in entity data and relationships into preset standard descriptive terms, and use the converted entity data and relationships as knowledge fragments;

[0023] When medical data is in the form of image data, semantic information is extracted using a multimodal image-text model.

[0024] The descriptive words in the semantic information are converted into preset standard descriptive words, and the converted semantic information is used as knowledge fragments.

[0025] In one implementation of this application, calculating the similarity value between the knowledge tag vector and the proprietary tag vector specifically includes:

[0026] The similarity value between knowledge tag vectors and proprietary tag vectors is calculated using cosine similarity.

[0027] In one implementation of this application, based on the corresponding knowledge fragment, actual retrieval needs, and a pre-set knowledge database, a large-scale medical AI model is used to obtain the output answer result, specifically including:

[0028] Input the corresponding knowledge fragments as the context of the actual retrieval needs into the medical AI big model;

[0029] The medical AI model outputs the answer results;

[0030] The answer also includes the knowledge snippets referenced in each sentence and the URLs of the corresponding medical data.

[0031] In one implementation of this application, after obtaining the output answer, the method further includes:

[0032] Automatically monitor and collect user-corrected and added data;

[0033] Based on the data types of the corrected and added data, knowledge fragments are extracted from them using the corresponding pre-trained large model.

[0034] Update the knowledge fragments to the preset knowledge database.

[0035] Secondly, this application provides a medical personal knowledge base management system based on a large model, the system comprising:

[0036] The storage module is used to determine the medical data acquisition interface based on the current user's personal knowledge needs, and acquire medical data from several sources; determine the corresponding pre-trained large model based on the specific type of medical data; use the pre-trained large model to extract knowledge fragments with preset standardized descriptions from the medical data, and store the knowledge fragments and the URLs of the corresponding medical data into a preset knowledge database;

[0037] The tag module is used to extract knowledge tags from knowledge fragments; collect multi-dimensional exclusive tags of the current user; among which, multi-dimensional exclusive tags include at least: professional field, interest topics, and behavioral preferences; vectorize the current user's knowledge tags and multi-dimensional exclusive tags to obtain knowledge tag and multi-dimensional exclusive tag vectors; add the knowledge tag vectors and the relationship between the knowledge tag vectors and knowledge fragments to the preset knowledge database;

[0038] The display module is used to calculate the similarity value between knowledge tag vectors and unique tag vectors; based on the similarity value, it determines the arrangement order and recommendation order of knowledge fragments in the current user's reading interface;

[0039] The answer module is used to configure a semantic question-answering framework in the reading interface. The actual search requirements are obtained through the input boxes in the semantic question-answering framework; the actual search requirements are analyzed by a large model to obtain semantic structure and keyword information; based on the relationship between keyword information and knowledge tag vectors, the corresponding knowledge fragments are determined from a preset knowledge database; based on the corresponding knowledge fragments, the actual search requirements, and the preset knowledge database, the output answer results are obtained using a large medical AI model.

[0040] In one implementation of this application, the storage module includes an acquisition unit.

[0041] It is used to input personal knowledge needs into a standardized description AI model to obtain retrieval information;

[0042] Input the search information into various medical data acquisition interfaces to obtain medical data from several sources that meet your personal knowledge needs.

[0043] Thirdly, this application provides a medical personal knowledge base management device based on a large model, the device comprising:

[0044] processor;

[0045] And a memory containing executable code, which, when executed, causes the processor to perform a large-model-based medical personal knowledge base management method, as described above.

[0046] Fourthly, this application provides a non-volatile computer storage medium storing computer instructions, which, when executed, implement a large-model-based medical personal knowledge base management method as described above.

[0047] As can be seen from the above technical solutions, this application has the following advantages:

[0048] I. Achieving standardized integration and intelligent modeling of cross-modal medical knowledge:

[0049] By using a pre-trained large-scale model to uniformly process multi-source heterogeneous medical data, this approach addresses the core pain point of traditional medical knowledge bases in integrating cross-modal data such as text, images, and clinical guidelines. Specifically, this manifests in: ① Flexible interface adaptation: Dynamically calling medical data interfaces from different sources based on user needs, ensuring comprehensive data acquisition; ② Knowledge structuring capability: Utilizing a large-scale model to extract standardized knowledge fragments and store associated URLs, preserving the original data traceability capability while constructing machine-understandable semantic knowledge units; ③ Automatic tagging: Extracting and quantifying tags from knowledge fragments, laying the data foundation for subsequent personalized recommendations.

[0050] II. Constructing a dynamic and personalized knowledge service mechanism:

[0051] Through multi-dimensional tag vector matching, a truly personalized knowledge service is achieved: ① Deep user profile modeling: Vectorization of multi-dimensional exclusive tags such as professional fields, topics of interest, and behavioral preferences can accurately depict the characteristics of users' knowledge needs; ② Intelligent sorting and recommendation: By calculating the similarity value between knowledge tag vectors and user-specific tag vectors, the order of knowledge fragments in the reading interface is dynamically adjusted, so that high-frequency knowledge is presented first; ③ Continuous optimization capability: The continuous update mechanism of the tag vector relationship library (such as automatically triggering tag matching when adding new knowledge fragments) ensures that the system adapts and evolves with changes in user behavior.

[0052] III. Provide intelligent question-answering services based on context understanding:

[0053] The semantic question-answering framework is designed to give the system clinical decision support-level retrieval capabilities: ① Deep demand analysis: The large model analyzes the semantic structure of natural language retrieval demands, which can accurately identify the user's implicit intent; ② Context-enhanced retrieval: By using related knowledge fragments as context and combining the reasoning ability of the medical AI large model, the answer results not only include directly matched content, but also generate clinical suggestions that integrate multiple knowledge points; ③ Knowledge traceability: All recommended answers are derived from standardized knowledge fragments in the preset knowledge database, and the original literature traceability path is retained through the associated URL, effectively avoiding the "illusion" risk of generative AI. Attached Figure Description

[0054] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of a medical personal knowledge base management method based on a large model, provided in an embodiment of this application.

[0056] Figure 2 This is a schematic diagram of the internal structure of a medical personal knowledge base management system based on a large model, provided in an embodiment of this application.

[0057] Figure 3 This is a schematic diagram of the internal structure of a medical personal knowledge base management device based on a large model, provided in an embodiment of this application. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Those skilled in the art should understand that the embodiments described below are merely preferred embodiments of this disclosure and do not imply that this disclosure can only be implemented through these preferred embodiments. These preferred embodiments are merely used to explain the technical principles of this disclosure and are not intended to limit the scope of protection of this disclosure. Based on the preferred embodiments provided by this disclosure, all other embodiments obtained by those skilled in the art without creative effort should still fall within the scope of protection of this disclosure.

[0060] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0061] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0062] The embodiment provides a method for managing a medical personal knowledge base based on a large model, such as Figure 1 As shown in the embodiments of this application, the method mainly includes the following steps:

[0063] Step 110: Based on the current user's personal knowledge needs, determine the medical data acquisition interface and acquire medical data from several sources; based on the specific type of medical data, determine the corresponding pre-trained large model; using the pre-trained large model, extract knowledge fragments with preset standardized descriptions from the medical data, and store the knowledge fragments and the URLs of the corresponding medical data into the preset knowledge database.

[0064] Understandably, the automated process design in this step achieves three technical optimizations: First, dynamically matching medical data interfaces and pre-trained models based on user needs avoids adaptation losses when general models process specialized data, improving the accuracy of knowledge extraction. Second, adopting a standardized knowledge fragment storage mechanism allows heterogeneous medical data from different sources (such as clinical guidelines, journal articles, case reports, etc.) to be stored in a unified structure, significantly reducing data cleaning costs during subsequent knowledge fusion. Finally, by retaining the original data URLs for traceability, it not only meets academic citation standards but also provides a traceable path for verifying knowledge credibility. This step, while maintaining data objectivity, improves the efficiency of the medical knowledge acquisition and storage process through technical architecture optimization.

[0065] In some embodiments, based on the user's current personal knowledge needs, the retrieval information for the medical data acquisition interface is determined, and medical data from several sources is acquired, specifically including:

[0066] Input your personal knowledge needs into a standardized AI model to obtain search information;

[0067] Input the search information into various medical data acquisition interfaces to obtain medical data from several sources that meet your personal knowledge needs.

[0068] Understandably, this approach uses standardized AI models to transform users' personalized knowledge needs into structured retrieval information, avoiding the subjective biases and efficiency bottlenecks of manually constructed search queries. This is particularly suitable for non-professional users to quickly and accurately express their needs. The parallel acquisition mechanism of multi-source medical data breaks the limitations of a single database, covering traditional academic resources such as journal articles and clinical guidelines, while also integrating practical data such as electronic medical records and health records to form a multi-dimensional knowledge graph. The system's output data strictly adheres to the objectivity of the original sources, without secondary processing or exaggerated interpretation, ensuring the authenticity and traceability of medical decision support information and providing a reliable foundation for subsequent professional analysis. The entire process achieves a technical closed loop of standardized demand transformation, comprehensive data collection, and transparent result output.

[0069] Specifically, based on the specific type of medical data, a corresponding pre-trained large model is determined; using the pre-trained large model, knowledge fragments with pre-defined standardized descriptions are extracted from the medical data, specifically including:

[0070] When medical data is in text format, a pre-trained large medical model is used to extract entity data and the relationships between entity data from the text data.

[0071] Convert the descriptive terms involved in entity data and relationships into preset standard descriptive terms, and use the converted entity data and relationships as knowledge fragments;

[0072] When medical data is in the form of image data, semantic information is extracted using a multimodal image-text model.

[0073] The descriptive words in the semantic information are converted into preset standard descriptive words, and the converted semantic information is used as knowledge fragments.

[0074] Understandably, by selecting appropriate pre-trained models based on data type, targeted processing of text and image data is ensured. For text data, the medical pre-trained model effectively identifies entities and relationships, and eliminates terminological differences through standardized descriptive term conversion, ensuring the uniformity of knowledge fragments. For image data, the multimodal model accurately extracts semantic information and then performs standardized conversion, transforming visual information into structured knowledge units. This categorized processing mechanism preserves the professional details of the original data while achieving cross-source data format uniformity through pre-set standard descriptive terms, providing standardized input for subsequent knowledge integration and retrieval, and avoiding information loss or bias caused by manual annotation or single-model processing. The entire process achieves a reliable transformation of medical data from its raw form to standardized knowledge fragments.

[0075] Step 120: Extract knowledge tags from knowledge fragments; collect multi-dimensional exclusive tags for the current user; vectorize the knowledge tags and multi-dimensional exclusive tags for the current user to obtain knowledge tag and multi-dimensional exclusive tag vectors; add the knowledge tag vectors and the relationship between the knowledge tag vectors and knowledge fragments to the preset knowledge database.

[0076] Among them, the multi-dimensional exclusive tags include at least: professional field, focus of attention, and behavioral preferences.

[0077] Understandably, this step, by extracting knowledge tags from knowledge fragments and combining them with users' multi-dimensional, personalized tags (such as professional fields, topics of interest, and behavioral preferences), can construct a more accurate user profile, making recommended content more tailored to individual needs. Secondly, tag vectorization transforms discrete tag information into a computable mathematical expression, facilitating subsequent similarity matching and knowledge association analysis, thus improving retrieval efficiency. Finally, storing the knowledge tag vectors and their relationships with knowledge fragments in a pre-defined knowledge database not only achieves structured knowledge storage but also supports dynamic updates, enabling the system to continuously adapt to changes in user preferences.

[0078] Step 130: Calculate the similarity value between the knowledge tag vector and the exclusive tag vector; based on the similarity value, determine the arrangement order and recommendation order of knowledge fragments in the current user's reading interface.

[0079] The calculation of the similarity value between the knowledge tag vector and the specific tag vector includes:

[0080] The similarity value between knowledge tag vectors and proprietary tag vectors is calculated using cosine similarity.

[0081] Understandably, this step, by calculating the cosine similarity between knowledge tag vectors and user-specific tag vectors, objectively quantifies the degree of matching between knowledge content and user characteristics, avoiding the arbitrariness of subjective ranking. Similarity value calculation provides an interpretable mathematical basis for the personalized ranking of knowledge fragments, allowing the reading interface to prioritize content that highly aligns with the user's expertise, interests, and behavioral preferences. This ranking method retains the transparency of the recommendation system (each ranking result is supported by a corresponding similarity value) while adapting to the dynamic changes in different users' needs. When user tags are updated, the system can quickly recalculate and adjust the display order. The entire process achieves an automated mapping from user characteristics to knowledge presentation, maintaining the relevance of recommendations without manual intervention.

[0082] Step 140: Configure the semantic question-answering framework in the reading interface. Obtain the actual search requirements through the input boxes in the semantic question-answering framework. Analyze the actual search requirements through the large model to obtain semantic structure and keyword information. Determine the corresponding knowledge fragments from the preset knowledge database based on the relationship between keyword information and knowledge tag vectors. Based on the corresponding knowledge fragments, actual search requirements, and preset knowledge database, use the medical AI large model to obtain the output answer results.

[0083] Understandably, this step, through a semantic question-answering framework, allows users to directly input their search requests in natural language, lowering the barrier to accessing professional medical knowledge. The large-scale model's analysis of search requests accurately captures semantic structure and keywords, avoiding missed detections caused by terminology differences in traditional keyword searches. Based on the association and matching of knowledge tag vectors and keywords, the system can quickly locate the most relevant knowledge fragments in the preset knowledge base, ensuring the professionalism and relevance of the search results. The medical AI large-scale model, combining the original knowledge fragments with the user's actual needs, generates answers that maintain the accuracy of medical information while adjusting the expression according to specific questions, making the output both professionally standardized and easy to understand. The entire process automates the transformation from vague requests to precise answers, reducing the time cost of manual retrieval and organization.

[0084] In some embodiments, based on corresponding knowledge fragments, actual retrieval needs, and a pre-set knowledge database, a large-scale medical AI model is used to obtain the output answer results, specifically including:

[0085] Input the corresponding knowledge fragments as the context of the actual retrieval needs into the medical AI big model;

[0086] The medical AI model outputs the answer results;

[0087] The answer also includes the knowledge snippets referenced in each sentence and the URLs of the corresponding medical data.

[0088] It's important to note that by using matched knowledge fragments as contextual input into the medical AI model, the generated answers are strictly confined to verified professional knowledge, preventing the model from generating misleading responses. The answers include sentence-level citations of knowledge fragments and URLs of the original data, providing triple protection: first, ensuring that every medical assertion is supported by authoritative sources, allowing users to trace back to original literature or clinical data; second, allowing professional users to verify the accuracy of information, meeting the requirements of evidence-based medicine; and third, distinguishing automatically generated content from original knowledge fragments through explicit annotation of citation relationships, adhering to the transparency principle of medical AI. This mechanism effectively controls the legal risks and ethical issues of medical information dissemination while maintaining the advantages of the large model's language generation capabilities.

[0089] After obtaining the output answer, the method also includes:

[0090] Automatically monitor and collect user-corrected and added data;

[0091] Based on the data types of the corrected and added data, knowledge fragments are extracted from them using the corresponding pre-trained large model.

[0092] Update the knowledge fragments to the preset knowledge database.

[0093] It's important to note that this system continuously captures new evidence and expert experience from clinical practice by automatically monitoring user corrections and supplementary data. Adaptive pre-trained models are used for knowledge extraction based on different data types, ensuring accurate entity relationship identification for text-based corrections (such as updated treatment plans) and effective feature extraction for image-based supplements (such as novel medical images). Updated knowledge fragments are immediately integrated into a pre-set database, forming a closed-loop iterative mechanism: on the one hand, it corrects outdated or erroneous information in the original knowledge base (such as adjusting drug contraindications according to the latest clinical guidelines); on the other hand, it expands the scope of knowledge coverage (such as adding treatment plans for rare diseases). This dynamic update mechanism keeps the system in sync with the forefront of medicine, while retaining user correction trajectories to provide audit clues for subsequent knowledge verification and quality control. The entire process enables the knowledge base to self-evolve, completing knowledge updates without service interruption.

[0094] As described above, this embodiment uses a pre-trained large model to uniformly process multi-source heterogeneous medical data, solving the core pain point of traditional medical knowledge bases' difficulty in integrating cross-modal data such as text, images, and clinical guidelines. Specifically, this manifests in: ① Flexible interface adaptation: Dynamically calling medical data interfaces from different sources based on user needs, ensuring comprehensive data acquisition; ② Knowledge structuring capability: Utilizing a large model to extract standardized knowledge fragments and store associated URLs, preserving the original data traceability capability while constructing machine-understandable semantic knowledge units; ③ Automatic tagging processing: Extracting and quantifying tags from knowledge fragments, laying the data foundation for subsequent personalized recommendations.

[0095] Through multi-dimensional tag vector matching, a truly personalized knowledge service is achieved: ① Deep user profile modeling: Vectorization of multi-dimensional exclusive tags such as professional fields, topics of interest, and behavioral preferences can accurately depict the characteristics of users' knowledge needs; ② Intelligent sorting and recommendation: By calculating the similarity value between knowledge tag vectors and user-specific tag vectors, the order of knowledge fragments in the reading interface is dynamically adjusted, so that high-frequency knowledge is presented first; ③ Continuous optimization capability: The continuous update mechanism of the tag vector relationship library (such as automatically triggering tag matching when adding new knowledge fragments) ensures that the system adapts and evolves with changes in user behavior.

[0096] The semantic question-answering framework is designed to give the system clinical decision support-level retrieval capabilities: ① Deep demand analysis: The large model analyzes the semantic structure of natural language retrieval demands, which can accurately identify the user's implicit intent; ② Context-enhanced retrieval: By using related knowledge fragments as context and combining the reasoning ability of the medical AI large model, the answer results not only include directly matched content, but also generate clinical suggestions that integrate multiple knowledge points; ③ Knowledge traceability: All recommended answers are derived from standardized knowledge fragments in the preset knowledge database, and the original literature traceability path is retained through the associated URL, effectively avoiding the "illusion" risk of generative AI.

[0097] In addition, this application Figure 2 This application provides a medical personal knowledge base management system based on a large model. For example... Figure 2 As shown in the embodiments of this application, the system mainly includes:

[0098] Storage module 210 is used to determine the medical data acquisition interface based on the current user's personal knowledge needs, acquire medical data from several sources; determine the corresponding pre-trained large model based on the specific type of medical data; use the pre-trained large model to extract knowledge fragments with preset standardized descriptions from the medical data, and store the knowledge fragments and the URLs of the corresponding medical data into a preset knowledge database.

[0099] Storage module 210 includes an acquisition unit,

[0100] It is used to input personal knowledge needs into a standardized description AI model to obtain retrieval information;

[0101] Input the search information into various medical data acquisition interfaces to obtain medical data from several sources that meet your personal knowledge needs.

[0102] Understandably, the automated process design of storage module 210 achieves three technical optimizations: First, it dynamically matches medical data interfaces and pre-trained models based on user needs, avoiding adaptation losses when general models process specialized data and improving the accuracy of knowledge extraction. Second, it adopts a standardized knowledge fragment storage mechanism, enabling heterogeneous medical data from different sources (such as clinical guidelines, journal articles, case reports, etc.) to be stored in a unified structured manner, significantly reducing data cleaning costs during subsequent knowledge fusion. Finally, by retaining the original data URLs for traceability, it not only meets academic citation standards but also provides a traceable path for knowledge credibility verification. This step, while maintaining data objectivity, improves the efficiency of the medical knowledge acquisition and storage process through technical architecture optimization.

[0103] The tag module 220 is used to extract knowledge tags from knowledge fragments; collect multi-dimensional exclusive tags of the current user; wherein, the multi-dimensional exclusive tags include at least: professional field, interest topics, and behavioral preferences; vectorize the current user's knowledge tags and multi-dimensional exclusive tags to obtain knowledge tag and multi-dimensional exclusive tag vectors; and add the knowledge tag vectors and the relationship between the knowledge tag vectors and knowledge fragments to the preset knowledge database.

[0104] Understandably, the tag module 220, by extracting knowledge tags from knowledge fragments and combining them with users' multi-dimensional exclusive tags (such as professional fields, topics of interest, and behavioral preferences), can construct more accurate user profiles, making recommended content more tailored to individual needs. Secondly, tag vectorization transforms discrete tag information into computable mathematical expressions, facilitating subsequent similarity matching and knowledge association analysis, thus improving retrieval efficiency. Finally, storing the knowledge tag vectors and their relationships with knowledge fragments in a pre-defined knowledge database not only achieves structured knowledge storage but also supports dynamic updates, enabling the system to continuously adapt to changes in user preferences.

[0105] The display module 230 is used to calculate the similarity value between the knowledge tag vector and the exclusive tag vector; based on the similarity value, it determines the arrangement order and recommendation order of knowledge fragments in the current user's reading interface.

[0106] Understandably, the display module 230 objectively quantifies the matching degree between knowledge content and user characteristics by calculating the cosine similarity between knowledge tag vectors and user-specific tag vectors, avoiding the arbitrariness of subjective ranking. Similarity value calculation provides an interpretable mathematical basis for the personalized ranking of knowledge fragments, allowing the reading interface to prioritize content that highly aligns with the user's expertise, interests, and behavioral preferences. This ranking method retains the transparency of the recommendation system (each ranking result is supported by a corresponding similarity value) while adapting to the dynamic changes in different users' needs. When user tags are updated, the system can quickly recalculate and adjust the display order. The entire process achieves an automated mapping from user characteristics to knowledge presentation, maintaining the relevance of recommendations without manual intervention.

[0107] The answer module 240 is used to configure a semantic question-answering framework in the reading interface. The actual search requirements are obtained through the input boxes in the semantic question-answering framework; the actual search requirements are analyzed by a large model to obtain semantic structure and keyword information; the corresponding knowledge fragments are determined from the preset knowledge database based on the relationship between keyword information and knowledge tag vectors; and the output answer results are obtained using the medical AI large model based on the corresponding knowledge fragments, actual search requirements and preset knowledge database.

[0108] Understandably, the answer module 240, through a semantic question-answering framework, allows users to directly input their search requests in natural language, lowering the barrier to accessing professional medical knowledge. The large-scale model's analysis of search requests accurately captures semantic structure and keywords, avoiding the missed detections caused by terminology differences in traditional keyword searches. Based on the association and matching of knowledge tag vectors and keywords, the system can quickly locate the most relevant knowledge fragments in the preset knowledge base, ensuring the professionalism and relevance of the search results. The medical AI large-scale model, combining the original knowledge fragments with the user's actual needs, generates answers that maintain the accuracy of medical information while adjusting the expression according to specific questions, making the output results both professionally standardized and easy to understand. The entire process automates the transformation from vague requests to precise answers, reducing the time cost of manual retrieval and organization.

[0109] As described above, this embodiment uses a pre-trained large model to uniformly process multi-source heterogeneous medical data, solving the core pain point of traditional medical knowledge bases' difficulty in integrating cross-modal data such as text, images, and clinical guidelines. Specifically, this manifests in: ① Flexible interface adaptation: Dynamically calling medical data interfaces from different sources based on user needs, ensuring comprehensive data acquisition; ② Knowledge structuring capability: Utilizing a large model to extract standardized knowledge fragments and store associated URLs, preserving the original data traceability capability while constructing machine-understandable semantic knowledge units; ③ Automatic tagging processing: Extracting and quantifying tags from knowledge fragments, laying the data foundation for subsequent personalized recommendations.

[0110] Through multi-dimensional tag vector matching, a truly personalized knowledge service is achieved: ① Deep user profile modeling: Vectorization of multi-dimensional exclusive tags such as professional fields, topics of interest, and behavioral preferences can accurately depict the characteristics of users' knowledge needs; ② Intelligent sorting and recommendation: By calculating the similarity value between knowledge tag vectors and user-specific tag vectors, the order of knowledge fragments in the reading interface is dynamically adjusted, so that high-frequency knowledge is presented first; ③ Continuous optimization capability: The continuous update mechanism of the tag vector relationship library (such as automatically triggering tag matching when adding new knowledge fragments) ensures that the system adapts and evolves with changes in user behavior.

[0111] The semantic question-answering framework is designed to give the system clinical decision support-level retrieval capabilities: ① Deep demand analysis: The large model analyzes the semantic structure of natural language retrieval demands, which can accurately identify the user's implicit intent; ② Context-enhanced retrieval: By using related knowledge fragments as context and combining the reasoning ability of the medical AI large model, the answer results not only include directly matched content, but also generate clinical suggestions that integrate multiple knowledge points; ③ Knowledge traceability: All recommended answers are derived from standardized knowledge fragments in the preset knowledge database, and the original literature traceability path is retained through the associated URL, effectively avoiding the "illusion" risk of generative AI.

[0112] The above are method embodiments of this application. Based on the same inventive concept, embodiments of this application also provide a medical personal knowledge base management device based on a large model. Figure 3 As shown, the device includes: a processor; and a memory storing executable code thereon, which, when executed, causes the processor to perform a large-model-based medical personal knowledge base management method as described in the above embodiments.

[0113] Specifically, the server determines the medical data acquisition interface based on the current user's personal knowledge needs, and acquires medical data from several sources; based on the specific type of medical data, it determines the corresponding pre-trained large model; using the pre-trained large model, it extracts knowledge fragments with pre-defined standardized descriptions from the medical data, and stores the knowledge fragments and the corresponding medical data URLs in a pre-defined knowledge database; it extracts knowledge tags from the knowledge fragments; it collects the current user's multi-dimensional exclusive tags; among which, the multi-dimensional exclusive tags include at least: specialty field, topics of interest, and behavioral preferences; it vectorizes the current user's knowledge tags and multi-dimensional exclusive tags to obtain knowledge tag and multi-dimensional exclusive tag vectors; and it then... The relationships between vectors, knowledge tag vectors, and knowledge fragments are added to a pre-defined knowledge database; similarity values ​​are calculated between knowledge tag vectors and exclusive tag vectors; based on the similarity values, the arrangement and recommendation order of knowledge fragments in the current user's reading interface are determined; a semantic question-answering framework is configured in the reading interface, and the actual search requirements are obtained through the input boxes in the semantic question-answering framework; the actual search requirements are analyzed through a large model to obtain semantic structure and keyword information; based on the relationship between keyword information and knowledge tag vectors, the corresponding knowledge fragments are determined from the pre-defined knowledge database; based on the corresponding knowledge fragments, the actual search requirements, and the pre-defined knowledge database, the output answer results are obtained using a large medical AI model.

[0114] In addition, this application embodiment also provides a non-volatile computer storage medium storing executable instructions, which, when executed, implement the above-described method for managing a large-scale medical personal knowledge base.

[0115] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for managing a medical personal knowledge base based on a large model, characterized in that, The method includes: Based on the current user's individual knowledge needs, a medical data acquisition interface is determined to acquire medical data from several sources; based on the specific type of medical data, a corresponding pre-trained large-scale model is determined; using the pre-trained large-scale model, knowledge fragments with pre-defined standardized descriptions are extracted from the medical data, specifically including: When medical data is text-based, a pre-trained medical model is used to extract entity data and the relationships between entities. The descriptive terms related to the entity data and relationships are converted into preset standard descriptive terms, and the converted entity data and relationships are used as knowledge fragments. When medical data is image-based, a multimodal image-text model is used to extract semantic information. The descriptive terms in the semantic information are converted into preset standard descriptive terms, and the converted semantic information is used as knowledge fragments. The knowledge fragments and the corresponding URLs of the medical data are stored in a preset knowledge database. Extract knowledge tags from knowledge fragments; collect multi-dimensional exclusive tags for the current user; among which, multi-dimensional exclusive tags include at least: professional field, topics of interest, and behavioral preferences; vectorize the current user's knowledge tags and multi-dimensional exclusive tags to obtain knowledge tag and multi-dimensional exclusive tag vectors; add the knowledge tag vectors and the relationship between the knowledge tag vectors and knowledge fragments to the preset knowledge database; Calculate the similarity value between the knowledge tag vector and the personal tag vector; based on the similarity value, determine the arrangement order and recommendation order of knowledge fragments in the current user's reading interface; Configure a semantic question-answering framework in the reading interface, obtain the actual search requirements through the input box in the semantic question-answering framework; analyze the actual search requirements through a large model to obtain semantic structure and keyword information; determine the corresponding knowledge fragments from the preset knowledge database based on the relationship between keyword information and knowledge tag vectors; and obtain the output answer results using the corresponding knowledge fragments, actual search requirements, and preset knowledge database using a large medical AI model. Automatically monitor user-corrected and added data; extract knowledge fragments from the corrected and added data using the corresponding pre-trained large model based on the data type; and update the knowledge fragments to the preset knowledge database.

2. The medical personal knowledge base management method based on a large model according to claim 1, characterized in that, Based on the user's current knowledge needs, determine the retrieval information for the medical data acquisition interface, and acquire medical data from several sources, specifically including: Input your personal knowledge needs into a standardized AI model to obtain search information; Input the search information into various medical data acquisition interfaces to obtain medical data from several sources that meet your personal knowledge needs.

3. The medical personal knowledge base management method based on a large model according to claim 1, characterized in that, Calculate the similarity value between the knowledge tag vector and the proprietary tag vector, specifically including: The similarity value between knowledge tag vectors and proprietary tag vectors is calculated using cosine similarity.

4. The medical personal knowledge base management method based on a large model according to claim 1, characterized in that, Based on the corresponding knowledge fragments, actual retrieval needs, and a pre-set knowledge database, a large-scale medical AI model is used to obtain the output answer results, specifically including: Input the corresponding knowledge fragments as the context of the actual retrieval needs into the medical AI big model; The medical AI model outputs the answer results; The answer also includes the knowledge snippets referenced in each sentence and the URLs of the corresponding medical data.

5. A medical personal knowledge base management system based on a large model, characterized in that, The system includes: The storage module is used to determine the medical data acquisition interface based on the user's current personal knowledge needs, and acquire medical data from several sources; based on the specific type of medical data, determine the corresponding pre-trained large model; and use the pre-trained large model to extract knowledge fragments with preset standardized descriptions from the medical data. Specifically, this includes: when the medical data is text-based, using the medical pre-trained large model to extract entity data and the relationships between entity data from the text data; converting the descriptive words involved in the entity data and relationships into preset standard descriptive words, and using the converted entity data and relationships as knowledge fragments; when the medical data is image data, using a text-image multimodal model to extract semantic information; converting the descriptive words in the semantic information into preset standard descriptive words, and using the converted semantic information as knowledge fragments; and storing the knowledge fragments and the corresponding URLs of the medical data in a preset knowledge database. The tag module is used to extract knowledge tags from knowledge fragments; collect multi-dimensional exclusive tags of the current user; among which, multi-dimensional exclusive tags include at least: professional field, interest topics, and behavioral preferences; vectorize the current user's knowledge tags and multi-dimensional exclusive tags to obtain knowledge tag and multi-dimensional exclusive tag vectors; add the knowledge tag vectors and the relationship between the knowledge tag vectors and knowledge fragments to the preset knowledge database; The display module is used to calculate the similarity value between knowledge tag vectors and unique tag vectors; based on the similarity value, it determines the arrangement order and recommendation order of knowledge fragments in the current user's reading interface; The answer module is used to configure a semantic question-answering framework in the reading interface. It obtains the actual search requirements through input boxes within the framework; it then analyzes these requirements using a large model to acquire semantic structure and keyword information; based on the relationship between keyword information and knowledge tag vectors, it determines the corresponding knowledge fragments from a pre-set knowledge database; and based on the corresponding knowledge fragments, the actual search requirements, and the pre-set knowledge database, it uses a large medical AI model to obtain the output answer; it automatically monitors user-corrected and added data; based on the data type of the corrected and added data, it extracts knowledge fragments from them using the corresponding pre-trained large model; and finally, it updates the knowledge fragments to the pre-set knowledge database.

6. The medical personal knowledge base management system based on a large model according to claim 5, characterized in that, The storage module includes an acquisition unit. It is used to input personal knowledge needs into a standardized description AI model to obtain retrieval information; Input the search information into various medical data acquisition interfaces to obtain medical data from several sources that meet your personal knowledge needs.

7. A medical personal knowledge base management device based on a large model, characterized in that, The device includes: processor; And a memory having executable code stored thereon, which, when executed, causes the processor to perform a large-model-based medical personal knowledge base management method as described in any one of claims 1-4.

8. A non-volatile computer storage medium, characterized in that, It stores computer instructions, which, when executed, implement a medical personal knowledge base management method based on a large model as described in any one of claims 1-4.

Images