Medical science popularization article recommendation method and system based on user portrait

Abstract

The invention provides a medical science popularization article recommendation method and system based on a user portrait. According to the method, multi-source medical data are integrated, logic association of the multi-source medical data is analyzed, a dynamic knowledge graph is constructed, then in the graph, real-time diagnosis and treatment behaviors of a user and historical medical interaction records are fused to generate a user portrait label set layered according to the medical cognition level, and the understanding depth and preference of the user on medical concepts are reflected; and further migrating preference features in the portrait from a non-real scene of historical interaction to a real scene of real-time diagnosis and treatment, and generating a dynamic recommendation feature vector synchronized with medical knowledge evolution through cross-scene conversion. And finally, according to a matching result of the feature vectors and concepts in the knowledge graph, dynamically adjusting an article recommendation priority, and generating a personalized recommendation sequence fitting the real-time cognitive level of the user. According to the technical scheme provided by the invention, the accuracy and timeliness of medical science popularization article recommendation can be improved.

Description

Technical Field

[0001] This application relates to the technical field of medical science popularization, and particularly to a method and system for recommending medical science popularization articles based on user portraits. Background Art

[0002] With the improvement of health awareness, users' demand for multi-dimensional medical science popularization knowledge such as disease prevention and medication guidelines is increasing day by day. Due to the significant differences among user groups in terms of health status, knowledge level, information preference, etc., the traditional "one-size-fits-all" popular science content distribution model is difficult to meet personalized needs.

[0003] The current technical solution is to use machine learning algorithms to classify a large number of medical science popularization articles and perform matching recommendations based on the feature data in the user portrait. This method first analyzes the content of medical science popularization articles through natural language processing technology, extracts keyword vocabulary and topic information, and then uses a machine learning model to perform intelligent matching according to the attributes in the user portrait to determine which articles are most likely to meet the needs of specific users. This method reduces manual intervention through an automated process, improves the recommendation efficiency and accuracy, and provides a more personalized reading experience for users.

[0004] However, the above existing solutions have certain defects. On the one hand, due to the complexity and professionalism of the medical field, natural language processing technology may encounter problems of insufficient understanding depth when analyzing medical science popularization articles, resulting in a deviation between the recommended articles and the actual needs of users; on the other hand, the construction of user portraits depends on multiple data sources. If the data collection is incomplete or not updated in a timely manner, it will directly affect the accuracy and relevance of the recommendation results. In addition, privacy protection is also an issue that cannot be ignored in the implementation process of this solution. How to protect the security of users' personal information while ensuring the recommendation effect is an important challenge faced by this solution. Summary of the Invention

[0005] This application provides a method and system for recommending medical science popularization articles based on user portraits to solve the problems of poor accuracy and timeliness in the recommendation of medical science popularization articles in the prior art.

[0006] In a first aspect, this application provides a method for recommending medical science popularization articles based on user portraits, including: By analyzing the logical relationship between unstructured text and structured case data in multi-source heterogeneous medical data, constructing a dynamic knowledge graph containing the dynamic dependence relationship between medical concepts; In the dynamic knowledge graph, fusing the obtained user real-time diagnosis and treatment behavior data and user historical medical interaction records with the medical concepts in multiple dimensions to generate a set of user portrait tags with the user's medical cognitive level as the stratification benchmark; Perform cross-scenario feature transformation on the user medical preference features in the user portrait tag set from the unrealistic scenarios in the user's historical medical interaction records to the realistic scenarios in the user's real-time diagnosis and treatment behavior data, and generate a dynamic recommendation feature vector that is updated synchronously with the evolution trend of medical knowledge; According to the joint matching result between the dynamic recommendation feature vector and the medical concepts in the dynamic knowledge graph, dynamically adjust the recommended priority ranking of medical popular science articles, and generate a personalized recommendation sequence that matches the user's real-time medical cognitive level.

[0007] Optionally, the performing cross-scenario feature transformation on the user medical preference features in the user portrait tag set from the unrealistic scenarios in the user's historical medical interaction records to the realistic scenarios in the user's real-time diagnosis and treatment behavior data, and generating a dynamic recommendation feature vector that is updated synchronously with the evolution trend of medical knowledge, includes: Extract the preference components of the unrealistic scenarios in the user's historical medical interaction records according to the user medical preference features in the user portrait tag set; Construct a non-linear projection relationship from the unrealistic scenarios to the realistic scenarios by analyzing the implicit association pattern between the operation trajectories of medical concepts in the unrealistic scenarios and the user's real-time diagnosis and treatment behavior data in the realistic scenarios; Based on the non-linear projection relationship, map the user medical preference features through the preference components of the unrealistic scenarios to the feature space of the realistic scenarios, and separate the scenario noise components irrelevant to the user's medical cognitive level in the feature space, and retain the core preference components that remain stable in the cross-scenario feature transformation; By tracking the change of the update status of the medical concepts, calculate the dynamic offset of the core preference components in the evolution trend of medical knowledge, and superimpose the dynamic offset on the core preference components to generate a dynamic recommendation feature vector containing the evolution trend of medical knowledge.

[0008] Optionally, the calculating the dynamic offset of the core preference components in the evolution trend of medical knowledge by tracking the change of the update status of the medical concepts includes: Obtain the record of the change of the association relationship of the medical concepts in the dynamic knowledge graph, and extract the newly added dependency relationships and the invalidated dependency relationships between medical concepts to obtain the update status mark of the evolution trend of medical knowledge; Based on the update status mark, determine the positive offset direction corresponding to the newly added dependency relationship and the negative offset direction corresponding to the invalidated dependency relationship in the core preference components; Based on the superposition effect of the positive offset direction and the negative offset direction, and combining the proportional weights of the new dependency relationship and the expired dependency relationship, the dynamic offset amount of the core preference component in the evolution trend of medical knowledge is obtained.

[0009] Optionally, determining the positive offset direction corresponding to the new dependency relationship and the negative offset direction corresponding to the expired dependency relationship in the core preference component based on the updated status flag includes: Based on the updated status flag, map the new dependency relationship to the first adjustment parameter of the core preference component, and map the expired dependency relationship to the second adjustment parameter of the core preference component; Generate a first set of direction vectors according to the product of the association strength between medical concepts connected by each new dependency relationship and the first adjustment parameter, and generate a second set of direction vectors according to the product of the historical association strength between medical concepts corresponding to each expired dependency relationship and the second adjustment parameter; Superimpose the direction vectors in the first set of direction vectors to obtain the positive offset direction, and superimpose the direction vectors in the second set of direction vectors to obtain the negative offset direction.

[0010] Optionally, in the dynamic knowledge graph, fusing the obtained user real-time diagnosis and treatment behavior data, user historical medical interaction records with the medical concepts in multiple dimensions to generate a set of user portrait tags based on the user's medical cognitive level as the stratification benchmark includes: Extract the operation timestamp and operation type identifier in the user real-time diagnosis and treatment behavior data, and the behavior frequency and behavior duration in the user historical medical interaction records; Align the operation timestamp with the time attribute of the medical concept in the dynamic knowledge graph in time series, and perform behavior matching according to the operation type identifier and the category attribute of the medical concept to obtain a preliminary association set; Calculate the stability weight of the association of medical concepts in the user historical medical interaction records according to the cumulative distribution ratio of the behavior frequency and behavior duration; Hierarchically superimpose the medical concepts in the preliminary association set with the stability weight, and combine the hierarchical distribution ratio of the medical concepts in the dynamic knowledge graph to generate a set of user portrait tags based on the user's medical cognitive level as the stratification benchmark.

[0011] Optionally, aligning the operation timestamp with the time attribute of the medical concept in the dynamic knowledge graph in time series, and performing behavior matching according to the operation type identifier and the category attribute of the medical concept to obtain a preliminary association set includes: Compare the operation timestamp with the time attribute of the medical concept in the dynamic knowledge graph. If the operation timestamp is within the effective time range of the time attribute, it is determined that the user's real-time diagnosis and treatment behavior is temporally aligned with the medical concept; Match the category association between the user's real-time diagnosis and treatment behavior and the medical concept according to the preset mapping relationship between the operation type identifier and the category attribute of the medical concept; Combine the user's real-time diagnosis and treatment behavior that is temporally aligned and has a matching category association with the medical concept into a behavior-concept association pair, and summarize each behavior-concept association pair to obtain a preliminary association set.

[0012] Optionally, the dynamically adjusting the recommended priority ranking of medical popular science articles according to the joint matching result between the dynamic recommendation feature vector and the medical concept in the dynamic knowledge graph, and generating a personalized recommendation sequence matching the user's real-time medical cognitive level includes: Calculate the matching degree of the association relationship between the dynamic recommendation feature vector and the medical concept in the dynamic knowledge graph, and extract the matching degree value directly associated with the medical concept in the dynamic recommendation feature vector; Set the priority adjustment rules for different medical concept levels according to the range of the matching degree value; Based on the priority adjustment rules, assign priority weights to the medical popular science articles associated with the user's real-time medical cognitive level in the dynamic knowledge graph; According to the priority weight assignment result, screen the medical popular science articles matching the user's real-time medical cognitive level from the dynamic knowledge graph, and perform the recommended priority ranking from high to low according to the priority weight assignment ratio to generate a personalized recommendation sequence.

[0013] In a second aspect, the present application provides a medical popular science article recommendation system based on a user portrait, including: An analysis module, by analyzing the logical association between unstructured text and structured case data in multi-source heterogeneous medical data, constructs a dynamic knowledge graph containing the dynamic dependence relationship between medical concepts; A fusion module, in the dynamic knowledge graph, fuses the obtained user's real-time diagnosis and treatment behavior data, the user's historical medical interaction records with the medical concept in multiple dimensions, and generates a set of user portrait labels with the user's medical cognitive level as the stratification benchmark; A conversion module, which performs cross-scene feature conversion on the user's medical preference features in the set of user portrait labels from the unrealistic scene of the user's historical medical interaction records to the realistic scene of the user's real-time diagnosis and treatment behavior data, and generates a dynamic recommendation feature vector that is synchronized with the evolution trend of medical knowledge; An adjustment module dynamically adjusts the recommended priority ranking of medical popular science articles according to the joint matching result between the dynamic recommendation feature vector and the medical concepts in the dynamic knowledge graph, and generates a personalized recommendation sequence that matches the user's real-time medical cognitive level.

[0014] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for recommending medical popular science articles based on a user portrait as described in the first aspect above.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it implements a method for recommending medical popular science articles based on a user portrait as described in the first aspect.

[0016] In the embodiment of the present application, by analyzing the logical association between unstructured text and structured case data in multi-source heterogeneous medical data, a dynamic knowledge graph containing dynamic dependency relationships between medical concepts is constructed, which can break through the semantic barriers of multi-source medical data, establish a medical concept relationship network that can be updated in real time, and provide an authoritative knowledge framework for accurate recommendation. In the dynamic knowledge graph, the obtained user real-time diagnosis and treatment behavior data and the user's historical medical interaction records are multi-dimensionally fused with the medical concepts to generate a set of user portrait tags with the user's medical cognitive level as the stratification benchmark. This step can fuse the user's real-time behavior and historical interaction data to realize the multi-dimensional quantitative expression of the user's needs. The user's medical preference features in the set of user portrait tags are subjected to cross-scene feature conversion from the unrealistic scenarios of the user's historical medical interaction records to the realistic scenarios of the user's real-time diagnosis and treatment behavior data, generating a dynamic recommendation feature vector that is updated synchronously with the evolution trend of medical knowledge. This step can eliminate the behavioral pattern differences between the virtual learning scenario and the real diagnosis and treatment scenario, transfer the user's preference features to the real demand scenario, and synchronize the dynamic evolution law of medical knowledge. According to the joint matching result between the dynamic recommendation feature vector and the medical concepts in the dynamic knowledge graph, the recommended priority ranking of medical popular science articles is dynamically adjusted, and a personalized recommendation sequence that matches the user's real-time medical cognitive level is generated. This step outputs a recommended result with dynamically adjusted content priority based on the intelligent matching between the user's real-time cognitive state and the knowledge graph, realizing the triple adaptation of needs, knowledge, and scenarios.

[0017] Furthermore, this method extracts the user's non-realistic scenario behavior features, explores their implicit associations with real-scenario diagnosis and treatment behaviors, and constructs a cross-scenario mapping model. By filtering out noise, the core cognitive features of the user are retained, and the dynamic offset is calculated in combination with the evolution direction of medical knowledge to generate a recommended feature vector that integrates the user's cognitive essence and the trend of knowledge update. This method solves the semantic deviation problem between virtual scenario behavior data and real demand scenarios, and through the dynamic offset mechanism, the recommended features can respond to the evolution of medical knowledge in real time, ensuring the accuracy and professional timeliness of the recommended content under scenario switching.

[0018] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 The flowchart of a method for recommending medical popular science articles based on user portraits provided by the present application is shown; Figure 2 The structural schematic diagram of a system for recommending medical popular science articles based on user portraits provided by the present application is shown; Figure 3 The structural schematic diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] In order to enable those skilled in the art to better understand the solutions of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application.

[0022] In some processes described in the specification, claims and above-mentioned drawings of the present application, there are multiple operations that appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this text or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this text are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0023] Researchers found that existing medical science popularization recommendation systems are difficult to dynamically integrate medical knowledge updates and user multi-scenario behavior differences, resulting in recommended content deviating from users' real-time cognitive needs and authoritative medical progress. Based on this, a method for recommending medical science popularization articles based on user portraits is provided. This method constructs a dynamic knowledge graph containing the dynamic dependence relationship between medical concepts, and combines the user's real-time diagnosis and treatment behavior data and historical medical interaction records to generate a set of tags reflecting the user's medical cognitive level and preferences. After cross-scenario feature transformation, a dynamic recommendation feature vector that is synchronized with the evolution trend of medical knowledge is formed. Finally, the recommendation priority is dynamically adjusted according to the matching result between this vector and medical concepts in the knowledge graph, achieving the goal of providing users with a personalized popular science article recommendation sequence that highly matches their current medical cognitive level. This method not only improves the relevance and practicality of information, but also promotes the effective dissemination of medical knowledge and the improvement of personal health management capabilities.

[0024] The technical solution of this application is applicable to the precise matching and recommendation scenario of medical science popularization articles for users' multi-dimensional health knowledge needs (such as disease prevention, medication guides). This method constructs a dynamic knowledge graph, combines users' real-time and historical medical data to generate personalized tags, and converts them into dynamic recommendation feature vectors. Then, according to the matching result of medical concepts, the recommendation priority of popular science articles is dynamically adjusted to achieve precise and personalized medical science popularization information recommendation, improve the relevance and practicality of information, and promote the effective dissemination of medical knowledge.

[0025] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0026] Figure 1 The following is a flowchart of a method for recommending medical science popularization articles based on user portraits provided for the embodiments of the present application. As Figure 1 shown, this method includes: 101. By analyzing the logical relationship between unstructured text and structured case data in multi-source heterogeneous medical data, construct a dynamic knowledge graph containing the dynamic dependence relationship between medical concepts; In this step, multi-source heterogeneous medical data refers to medical data sources in different formats such as electronic medical records, imaging reports, and scientific research literature.

[0027] Unstructured text includes free text data such as doctors' handwritten records and medical forum discussions, and structured case data refers to standardized data such as inspection indicators with standard codes and diagnostic codes.

[0028] Dynamic dependencies refer to the strength of associations between medical concepts such as disease and symptoms, drugs and side effects, which change with new research evidence.

[0029] The medical concept refers to structured case data, which is patient diagnosis and treatment information organized according to certain standards and formats.

[0030] A dynamic knowledge graph refers to a knowledge network built based on the logical associations between medical concepts (such as diseases and drugs), and its dependencies can be automatically adjusted as data is updated.

[0031] In the embodiment of the present application, the bidirectional long short-term memory network in the natural language processing technology is first used to perform entity recognition on the unstructured text to extract medical concepts such as disease names and drug ingredients. Then the diagnostic path of the structured case data is encoded using the graph attention network to generate weighted medical concepts. The medical concepts extracted from the two types of data sources are then logically associated through a dynamic graph convolutional network. When new case data is input, the node relationship weights in the graph structure are dynamically adjusted according to the time decay function and the strength of evidence. Finally, a dynamic knowledge graph containing time dimension features is formed.

[0032] When a tertiary hospital integrated the unstructured reports of the imaging department with the structured data of the laboratory department, the system discovered the "coronary artery calcification" entity through text analysis, and extracted the "elevated low-density lipoprotein" indicator from the structured test order, and established an initial association based on the international cardiovascular guidelines. When the newly released clinical research confirmed that the strength of the association between the two had changed, the dynamic knowledge graph automatically updated the edge weight value.

[0033] 102. In the dynamic knowledge graph, the acquired real-time diagnosis and treatment behavior data of the user, the historical medical interaction records of the user and the medical concepts are integrated in multiple dimensions to generate a user portrait label set based on the user's medical cognition level as a hierarchical basis; In this step, the user's real-time diagnosis and treatment behavior data refers to the user's operation sequence in the actual diagnosis and treatment scenario (such as prescription writing and test report query).

[0034] User historical medical interaction records refer to the user's behavior trajectory in non-realistic scenarios such as virtual learning and simulation training (such as disease simulation diagnosis exercises).

[0035] The user portrait label set refers to a group of labels that are hierarchically marked based on the level of medical cognition (such as elementary and advanced), reflecting the user's mastery of medical concepts.

[0036] In the embodiments of the present application, first, a medical concept trigger sequence (such as continuously querying "insulin dose adjustment" and "blood glucose monitoring frequency") is extracted from the user's real-time diagnosis and treatment behavior data through a behavior pattern extraction algorithm (such as time series pattern mining technology), and at the same time, a simulated operation trajectory sequence (such as the "diabetes complication screening path" in virtual diagnosis) is extracted from the user's historical medical interaction records. Then, the above behavior sequences are mapped to the corresponding nodes of the dynamic knowledge graph, and a hierarchical clustering algorithm (such as density-based spatial clustering method) is used to perform multi-dimensional fusion on the user operation frequency, concept complexity, and path depth to generate user medical cognitive level stratification labels (such as basic layer, advanced layer, expert layer). Finally, a user portrait label set is constructed according to the clustering results, where each label is associated with a specific medical concept cluster and its cognitive depth index in the graph.

[0037] Continuing with the above example, based on the latest diagnosis and treatment information and past medical history of each patient, they are classified in detail. For example, some patients may be more concerned about the management of chronic diseases, while others are interested in the emergency treatment of acute conditions. Based on this information, corresponding labels are assigned to different groups to form a detailed user portrait, which helps to better understand and meet the needs of patients.

[0038] 103. Perform cross-scenario feature transformation on the user medical preference features in the user portrait label set from the non-realistic scenario of the user's historical medical interaction records to the realistic scenario of the user's real-time diagnosis and treatment behavior data, and generate a dynamic recommendation feature vector that is updated synchronously with the evolution trend of medical knowledge; In this step, the non-realistic scenario refers to non-actual application environments such as virtual learning and simulated operations in the user's historical medical interaction records.

[0039] The realistic scenario refers to the actual demand scenarios such as real diagnosis and treatment and health management of the user's real-time diagnosis and treatment behavior.

[0040] Cross-scenario feature transformation refers to the process of migrating user preference features from a virtual environment to a real environment.

[0041] The dynamic recommendation feature vector refers to the mathematical representation that fuses user cognitive features and the evolution of medical knowledge.

[0042] In the embodiments of the present application, first, a feature decoupling algorithm (such as independent component analysis method) is used to separate non-realistic scenarios (such as the "cardiopulmonary resuscitation process" repeatedly operated in simulation training) from the user portrait tag set. Then, a cross-scenario non-linear mapping model (such as a feature transfer framework based on a generative adversarial network) is constructed, and cross-scenario feature transformation of non-realistic scenarios to realistic scenarios is performed through adversarial training technology, and an attention mechanism is used to filter out the noise generated by free exploration of the simulation environment (such as meaningless random jump operations). Finally, a knowledge evolution perception module (such as a temporal difference learning model) is introduced, and the dynamic offset of the core preference component is calculated according to the update intensity of medical concepts in the knowledge graph (such as the node weight change rate), and a dynamic recommendation feature vector that fuses the user's cognitive essence and the knowledge evolution trend is generated.

[0043] Based on previous cases, assume that a patient shows strong interest when reading an article about heart disease prevention. When this patient is admitted to the hospital due to chest pain, the system will take into account their previous interests and combine the latest heart disease treatment guidelines to provide personalized medical advice and support resources.

[0044] 104. Dynamically adjust the recommended priority ranking of medical popular science articles according to the joint matching result between the dynamic recommendation feature vector and medical concepts in the dynamic knowledge graph, and generate a personalized recommendation sequence that matches the user's real-time medical cognitive level.

[0045] In this step, the joint matching result refers to the correlation score between the dynamic recommendation feature vector and medical concepts in the knowledge graph.

[0046] The recommended priority ranking refers to the rule of dynamically adjusting the article recommendation order according to the matching degree.

[0047] The personalized recommendation sequence refers to a list of recommended items sorted by priority after analysis according to the user's interests, behavior habits and their current demand status. In the medical field, it refers to a series of medical popular science articles most suitable for the user.

[0048] In the embodiments of the present application, first, a multi-modal similarity calculation algorithm (such as an embedding vector matching method based on a graph neural network) is used to match the dynamic recommendation feature vector with the embedding representation of medical concept nodes in the dynamic knowledge graph to generate an initial correlation score. Then, based on a real-time feedback mechanism (such as an online learning to rank algorithm), the matching degree weight is dynamically adjusted according to implicit feedback data such as user clicks and dwell times, and the recommended priority ranking of medical popular science articles is recalculated by combining the initial correlation score and the incremental update result of the knowledge graph (such as the newly added "gene editing therapy" node). Finally, a personalized recommendation sequence is implemented through a streaming data processing engine to ensure that the recommendation sequence is strictly synchronized with the user's cognitive state and medical progress.

[0049] Taking the examples from the previous steps, for patients who showed an interest in heart disease prevention and were later admitted to the hospital due to chest pain, the system not only provided popular science articles related to heart disease but also adjusted the recommendation order in a timely manner according to their disease progression, giving priority to articles introducing first aid measures for acute myocardial infarction, greatly improving the relevance and practicality of the information.

[0050] In summary, steps 101 to 104 integrate the semantic associations of multi-source medical data through a dynamic knowledge graph, construct an authoritative knowledge framework that is updated in real time, generate a cognitive hierarchical portrait by fusing the user's multi-scenario behaviors, and accurately quantify the demand differences. Cross-scenario feature migration eliminates the behavior deviation between the virtual and real environments, and generates dynamic recommendation features by combining the knowledge evolution trend. Finally, it realizes the three-dimensional matching of medical popular science content with the user's cognitive state, real-time diagnosis and treatment needs, and medical progress, improving the accuracy, timeliness, and scenario adaptability of the recommendation.

[0051] To solve the semantic deviation problem between the user preferences in the virtual scenario and the real diagnosis and treatment needs, by analyzing the implicit association patterns between the operation trajectories of the user in the non-realistic scenario and the real-time diagnosis and treatment behavior data in the real scenario, a non-linear projection relationship is constructed to map these preference features to the feature space of the real scenario. Subsequently, by tracking the change of the updated state of medical concepts, the dynamic offset of the core preference component generated by the evolution trend of medical knowledge is calculated and superimposed on the core preference component, and finally a dynamic recommendation feature vector that can reflect the latest medical knowledge evolution is generated to achieve accurate and personalized medical information recommendation. This process ensures that the user's preference features can be effectively converted between different scenarios and keep synchronized with the latest medical knowledge. In some embodiments, the cross-scenario feature conversion of the user's medical preference features in the user portrait label set from the non-realistic scenario of the user's historical medical interaction records to the real scenario of the user's real-time diagnosis and treatment behavior data in step 103 to generate a dynamic recommendation feature vector that is synchronized with the evolution trend of medical knowledge includes: 201. Extract the preference components of the non-realistic scenario of the user's historical medical interaction records according to the user's medical preference features in the user portrait label set; In step 201, the preference components of the non-realistic scenario refer to the behavior characteristics generated by the user in non-real diagnosis and treatment scenarios such as simulation learning and virtual training, including the simulation operation path selection mode (such as the step jump sequence in the virtual diagnosis process), the dwell time distribution of knowledge nodes (such as the duration statistics of repeatedly consulting the mechanism of a certain disease), and the cross-medical concept jump logic chain (such as the coherent jump path from "symptom" to "treatment plan"). The user's medical preference features describe the user's special interests or tendencies in the medical field, such as the research on a certain disease and the attention to a specific therapy. These features are derived from the user's historical browsing records, search history, and other interaction behaviors.

[0052] In the embodiments of the present application, first, non-realistic scenario operation segments (such as the complete process records in simulation diagnosis and treatment exercises) are extracted from the user's historical medical interaction records through the behavior sequence segmentation technology. Secondly, a time series feature extraction model is used to encode the path selection and jump logic in the operation segments to generate a preference component vector of the user's medical preference features. Finally, the preference component vector is semantically aligned with the medical concept nodes in the dynamic knowledge graph (such as through a graph embedding matching algorithm) to form a preference component of the non-realistic scenario with semantic labels.

[0053] 202. By analyzing the implicit association pattern between the operation trajectory of the user on medical concepts in the non-realistic scenario and the real-time diagnosis and treatment behavior data of the user in the real scenario, a non-linear projection relationship from the non-realistic scenario to the real scenario is constructed; In step 202, the non-linear projection relationship refers to the mapping rule from the non-realistic scenario feature space to the real scenario feature space, which is established by quantifying the implicit logical association between the user's virtual operation trajectory (such as the simulation diagnosis path) and the actual diagnosis and treatment behavior (such as the prescription record) (such as the association strength between "antibiotic selection simulation" and "infection treatment prescription"). The implicit association pattern refers to the internal connection or law existing between the data. These connections are often not directly visible and need to be mined through data analysis techniques (such as machine learning algorithms). In cross-scenario feature transformation, this pattern helps to establish the mapping relationship between the non-realistic scenario and the real scenario.

[0054] In the embodiments of the present application, first, the implicit association pattern between the operation trajectory of the user on medical concepts in the non-realistic scenario and the real-time diagnosis and treatment behavior data is analyzed through machine learning algorithms (such as random forest or neural network). Secondly, a generator and discriminator framework is constructed according to the implicit association pattern. The generator maps the non-realistic scenario features to the real scenario space, and the discriminator optimizes the mapping rule by comparing the distribution differences between the generated features and the real behavior features. Finally, the non-linear projection relationship is iteratively adjusted through the adversarial training of the generator and discriminator framework, so that the mapped features approximate the real scenario distribution while retaining the user's cognitive essence.

[0055] 203. Based on the non-linear projection relationship, map the user's medical preference features to the feature space of the real scenario through the preference component of the non-realistic scenario, and separate the scenario noise component irrelevant to the user's medical cognitive level in the feature space, and retain the core preference component that remains stable and transmitted in cross-scenario feature transformation; In step 203, the scene noise component refers to the invalid features introduced by the degrees of freedom of unrealistic scene operations (such as random jumps and aimless browsing). The core preference component refers to the cognitive features stably transmitted by the user across scenes (such as the persistent attention to a specific treatment logic). The feature space refers to the multi-dimensional abstract space used to represent the characteristics of data, where each dimension represents a specific attribute or feature. In the medical field, the feature space can include information such as the age, gender, and medical history of patients, as well as more complex features extracted based on this information, such as disease risk scores. The user's medical cognitive level refers to the user's understanding and mastery of medical knowledge, including but not limited to the understanding of diseases, the understanding of treatment plans, and the basic common sense of health management. This level can be evaluated through various data such as the user's learning records and interaction behaviors.

[0056] In the embodiments of the present application, first, the mapped unrealistic scene features are input into the attention weight calculation model to analyze the contribution degrees of each feature dimension to the prediction of the feature space behavior of the realistic scene (such as the prediction weight of the "simulated path selection" feature for "actual prescription writing"). Secondly, based on the contribution degree threshold (such as the preset weight quantile), low-contribution features (such as random jump operations) are filtered, and high-contribution features are retained as the core preference component. Finally, the core preference component is embedded into the realistic scene feature space through the feature reconstruction technology to form the denoised core preference component.

[0057] 204. By tracking the change in the update status of the medical concept, calculate the dynamic offset of the core preference component in the evolution trend of medical knowledge, and superimpose the dynamic offset on the core preference component to generate a dynamic recommendation feature vector including the evolution trend of medical knowledge.

[0058] In step 204, the dynamic offset refers to the directional adjustment amount generated by the evolution of medical knowledge (such as the release of new diagnosis and treatment guidelines and the update of drug taboos) for the core preference component, which is used to synchronize the user's cognition with the update of authoritative knowledge. The evolution trend of medical knowledge refers to the change and development direction of knowledge in the medical field over time. This trend can be captured by tracking the change in the update status of medical concepts, reflecting the progress or adjustment of medical theory, practice, and technology. The dynamic recommendation feature vector is a data structure that combines the user's preferences and the latest development of medical knowledge, and is used in the personalized recommendation system to improve the relevance and timeliness of the recommended content.

[0059] In the embodiments of the present application, first, the update status of medical concepts is monitored, and text analysis techniques are used to track relevant changes. Then, the dynamic offset of the core preference component under the new trend is calculated, which usually requires combining time series analysis methods to evaluate the development trend of preferences over time. Finally, this dynamic offset is added to the original core preference component to generate a directional adjustment amount that includes the latest medical knowledge trend. In this way, it can be ensured that the health information recommended to users is always up-to-date and conforms to the current cognitive level and preferences of users.

[0060] The following is a specific example: In a certain top-three hospital, in order to improve the patient service experience and the accuracy of personalized medical information recommendation, the above method is adopted for optimization. First, by analyzing the browsing history, interaction records, and online course participation on the patient's health education platform, the interest preferences of patients for specific medical topics such as heart disease prevention are identified, and the preference intensity in each field is calculated to form a set of preference components. Next, machine learning algorithms are used to mine the potential connections between these preferences and the actual behavior data of patients during actual medical treatment. For example, analyzing the actual medical treatment behavior of patients after reading heart disease prevention materials, including the selection of physical examination items, drug use, etc., constructing a non-linear projection model to accurately map theoretical interests to actual medical treatment needs. On this basis, signal processing techniques such as principal component analysis are applied to distinguish stable core preference components from scenario noise components. By removing those fluctuation factors caused by specific situations, the core preference features that truly reflect the long-term interests and needs of patients are retained. Finally, combined with the latest medical research results, the update status of knowledge related to heart disease prevention is monitored, the development trend of core preference components over time is calculated, and the preference features of patients are dynamically adjusted. In this way, the system can provide personalized medical service recommendations for patients based on the latest medical evidence, such as recommending the latest heart disease prevention guidelines or treatment methods, ensuring that the recommended content is always up-to-date and conforms to the current cognitive level and preferences of patients.

[0061] In summary, steps 201 to 204 eliminate random noise in the virtual environment through cross-scenario feature transformation, and retain the essential features of user cognition; dynamically correct the recommendation direction in combination with the evolution trend of medical knowledge, and achieve the precise synchronization of the recommended content with the user's real-time medical treatment needs and the update of authoritative knowledge. Finally, it solves the semantic deviation problem caused by scenario fragmentation in traditional recommendation systems, improves the accuracy of cross-scenario recommendations, and ensures that the recommended results have both professional timeliness and personalized adaptation capabilities.

[0062] In order to further improve the calculation accuracy of the dynamic offset of the core preference component in the evolution trend of medical knowledge, an update status marking system is constructed by tracking the new and invalid records of the concept association relationships in the knowledge graph. Calculation models for the positive and negative offset directions are designed, and the impact of knowledge evolution on the user's core preference is quantified by combining proportional weights. This method innovatively transforms the knowledge evolution trend into a computable vector space movement, and realizes the dynamic calibration of the preference component through superposition, solving the problem of error accumulation caused by the lag of knowledge update in traditional recommendation systems. In some embodiments, calculating the dynamic offset of the core preference component in the evolution trend of medical knowledge as described in step 303 includes: 301. Obtain the change records of the association relationships of the medical concepts in the dynamic knowledge graph, extract the newly added dependency relationships and the invalidated dependency relationships between medical concepts, so as to obtain the update status marking of the medical knowledge evolution trend; In step 301, the association relationship change record refers to the time-series change dataset of the connection relationships between medical concept nodes in the dynamic knowledge graph, including attributes such as the relationship establishment time and the invalidation time. The newly added dependency relationship refers to the newly generated semantic association between concepts, and the invalidated dependency relationship refers to the old connection that is no longer valid after being verified by authority. The update status marking is an identifier or label used to reflect the change of the association relationship between specific medical concepts. This marking is usually generated based on the newly added dependency relationships and the invalidated dependency relationships between medical concepts recorded in the dynamic knowledge graph.

[0063] In the embodiments of the present application, first, the version management function of the dynamic knowledge graph is used to compare the graph structure snapshots in adjacent time periods to identify the change records of the association relationships between nodes. Then, through the graph structure difference analysis algorithm, the subgraph matching technology is used to scan the newly added semantic association edges, and at the same time, the expired invalid association edges are located based on the timestamp filtering mechanism. Finally, the detected newly added dependency relationships are marked as "positive evolution events", the invalidated dependency relationships are marked as "negative evolution events", and the time attributes and influence range parameters of each event are recorded to form an update status marking with the medical knowledge evolution trend.

[0064] 302. Based on the update status marking, determine the positive offset direction corresponding to the newly added dependency relationship and the negative offset direction corresponding to the invalidated dependency relationship in the core preference component; In step 302, the positive offset direction represents the vector of the enhancement effect of the semantic expansion of the medical concept due to the newly added association on the core preference. The negative offset direction represents the vector of the weakening effect of the concept dimension contraction caused by the invalidation of the old association on the core preference.

[0065] In the embodiments of the present application, first, the status flag is updated by applying graph representation learning technology to map medical concepts and their associated relationships into a low-dimensional vector space, so that concepts with similar semantics have proximity in the vector space. Then, for the newly added dependency relationship, the vector direction difference between the concept nodes at both ends of the relationship is calculated, and the change in cosine similarity is used as the positive offset direction. Finally, for the invalidated dependency relationship, the attenuation value of the association strength of the relationship in the historical vector model is extracted as the negative offset direction.

[0066] 303. Obtain the dynamic offset of the core preference component in the evolution trend of medical knowledge according to the superposition effect of the positive offset direction and the negative offset direction, and in combination with the proportional weights of the newly added dependency relationship and the invalidated dependency relationship.

[0067] In step 303, the proportional weights are comprehensively calculated from indicators such as the confidence levels, academic impact factors, and clinical verification levels of the newly added dependency relationship and the invalidated dependency relationship, and are used to quantify the contribution degrees of different change events to the offset. The dynamic offset refers to the adjustment value generated by the change of the user's preference characteristics according to new medical concepts, treatment methods, or health suggestions as medical knowledge is updated and developed.

[0068] In the embodiments of the present application, first, a multi-dimensional weight evaluation system is established, and characteristic data such as the credibility of the academic source, the number of clinical empirical evidences, and the time freshness of knowledge change events are collected, and the contribution degree weight coefficients of each characteristic are calculated by the entropy weight method. Then, the positive and negative offset vectors obtained in step 302 are respectively multiplied by the comprehensive weights of the corresponding events to realize the proportional weights of the influence of different evolution events. Finally, the vector space linear superposition principle is used to perform a synthesis operation on the weighted positive and negative vector groups, and the dimension difference is eliminated through normalization processing, and finally a standardized dynamic offset is output.

[0069] The following is a specific example: In a certain tertiary hospital, in the knowledge evolution scenario of the treatment field of cardiovascular diseases, the dynamic knowledge graph newly adds an association between "gene mutation and antiplatelet drug reactivity" (weight 0.8), and at the same time invalidates the old association between "β-blocker and asthma contraindication" (weight 0.6). Step 301 extracts these two change events and marks the status. Step 302 calculates the positive offset vector (0.12, 0.08) of the concepts related to gene therapy and the negative offset vector (-0.09, 0.05) of the respiratory system association respectively. Step 303 performs weighted superposition according to the weights, and finally obtains the dynamic offset (0.042, 0.094), guiding the recommendation system to adjust in the direction of precision medicine.

[0070] In summary, steps 301 to 303 effectively solve the problem of recommendation bias caused by lagging knowledge update in the medical field in traditional recommendation systems by establishing a dynamic mapping mechanism between knowledge evolution and user preferences. The system can automatically capture the fine-grained changes in medical concept relationships, and combine with a multi-dimensional weight evaluation system to accurately quantify the direction and intensity of the impact of knowledge evolution on users' core preferences. Compared with static models, this method significantly improves the adaptability of recommendation results to the latest medical progress and the filtering efficiency of obsolete knowledge, ensuring that the recommended content always conforms to the latest medical consensus while maintaining the stable evolution of personalized features. The graph difference analysis and vector space modeling methods adopted in the technical implementation process provide an interpretable computational framework for dealing with complex knowledge evolution.

[0071] To accurately quantify the dynamic impact of medical knowledge evolution on user preferences, a direction vector is generated by multiplying the association strength and the adjustment parameter, and the vector space superposition technology is used to synthesize the net impact direction. This solution break throughly transforms discrete knowledge change events into continuous vector operations, and uses the historical association strength to retain the attenuation effect of invalid relationships, ensuring that the preference adjustment responds to the latest medical progress and is compatible with historical cognitive inertia. In some embodiments, step 302 of determining the positive offset direction corresponding to the newly added dependency relationship and the negative offset direction corresponding to the invalidated dependency relationship in the core preference component based on the update status flag includes: 401. Based on the update status flag, map the newly added dependency relationship to the first adjustment parameter of the core preference component, and map the invalidated dependency relationship to the second adjustment parameter of the core preference component; In step 401, the first adjustment parameter is a quantization coefficient reflecting the degree of influence of the newly added dependency relationship on the core preference, including dimensions such as knowledge authority and evidence level. The second adjustment parameter is an attenuation coefficient characterizing the degree of weakening of the invalidated dependency relationship on the core preference, including elements such as historical citation frequency and failure duration.

[0072] In the embodiments of the present application, first, a knowledge evolution impact evaluation model is constructed through the update status flag. For the newly added dependency relationship, features such as the impact factor of the source journal (such as a journal coefficient of 5.2) and the number of multi-center clinical trials (such as 3 phase III trials) are extracted, and are normalized to the first adjustment parameter in the 0-1 interval through an S-shaped function. For the invalidated dependency relationship, the time decay curve of the number of citations in the last five years (such as an exponential decay factor of 0.3) is calculated, and combined with the failure confirmation duration (such as 18 months of invalidation), a linear regression model is used to generate the second adjustment parameter.

[0073] 402. Generate a first set of direction vectors based on the product of the association strength between medical concepts connected by each newly added dependency relationship and the first adjustment parameter, and generate a second set of direction vectors based on the product of the historical association strength between medical concepts corresponding to each invalidated dependency relationship and the second adjustment parameter. In step 402, the association strength refers to the tightness of the semantic connection between medical concepts, which is represented by the relationship weight of the knowledge graph. The historical association strength refers to the average influence of the invalidated relationship in the past period. A direction vector is a weighted multi-dimensional spatial displacement quantity, which characterizes the direction and amplitude of the adjustment of the preference component. The first set of direction vectors is a set composed of a series of vectors generated by multiplying the association strength between medical concepts connected by newly added dependency relationships in the dynamic knowledge graph by the corresponding adjustment parameter (i.e., the first adjustment parameter). The second set of direction vectors is a set composed of a series of vectors generated by multiplying the historical association strength between medical concepts corresponding to invalidated dependency relationships in the dynamic knowledge graph by the corresponding adjustment parameter (i.e., the second adjustment parameter).

[0074] In the embodiments of the present application, first, obtain the current weight value of the newly added relationship from the dynamic knowledge graph (such as the association strength between "gene detection and targeted drugs" is 0.78), and multiply it by the first adjustment parameter to obtain the first set of direction vectors. For the invalidated dependency relationship, call the historical version library to obtain its average weight in the three years before invalidation (such as the historical strength of "traditional chemotherapy and immunosuppression" is 0.65), and multiply it by the second adjustment parameter as the second set of direction vectors. Then, through the graph embedding projection technology, convert the scalar product into a displacement vector in the vector space, and retain the topological relationship characteristics of the concept nodes.

[0075] 403. Superimpose the direction vectors in the first set of direction vectors to obtain a positive offset direction, and superimpose the direction vectors in the second set of direction vectors to obtain a negative offset direction.

[0076] In step 403, vector superposition is to perform algebraic summation on the coordinate components of multiple direction vectors through the spatial vector synthesis rule. The positive offset direction is the net influence direction generated by the group of newly added relationships, and the negative offset direction is the net weakening direction caused by the group of invalidated relationships. The negative offset direction refers to the indication that due to the invalidation or weakening of the dependency relationship between some medical concepts, the user preference characteristics change in the direction of reducing interest or demand.

[0077] In the embodiments of the present application, first, all vectors in the first direction vector set are subjected to coordinate decomposition, and the axial components are obtained by superimposing and summing according to dimensions respectively (such as +0.34 on the X-axis and -0.12 on the Y-axis). The second direction vector set is processed in the same way to obtain the axial components (such as -0.25 on the X-axis and +0.08 on the Y-axis). Then, the magnitude normalization process of the vectors is used to eliminate the order-of-magnitude differences of the axial components, and finally, the positive offset direction vector (0.94, -0.33) and the negative offset direction vector (-0.95, 0.32) with unit length are formed.

[0078] The following is a specific example: In the scenario of updating diabetes diagnosis and treatment knowledge in a certain Class-III Grade-A hospital, the dynamic knowledge graph newly adds the association between "gut microbiota detection and insulin resistance" (authority coefficient 0.9, association strength 0.8), and invalidates the old association between "metformin and vitamin B12 deficiency" (historical strength 0.7, attenuation coefficient 0.6). In step 401, the first adjustment parameter 0.72 (0.9×0.8) and the second adjustment parameter 0.42 (0.7×0.6) are calculated. In step 402, the newly added relationship is mapped to the direction vector (0.58, 0.15), and the invalidated relationship is mapped to (-0.36, 0.22). After superimposing other relevant vectors in step 403, the positive offset direction (0.62, 0.31) and the negative offset direction (-0.55, 0.28) are finally generated, guiding the recommendation system to strengthen the push of content related to microbiome treatment and weaken the warning of outdated drug side effects at the same time.

[0079] In summary, steps 401 to 403 effectively solve the key problem that traditional recommendation systems are difficult to adapt to the dynamic evolution of medical knowledge by establishing a quantitative mapping mechanism between knowledge evolution events and the preference space. The system can accurately analyze the differential impacts of newly added knowledge and eliminated knowledge on user preferences, and uses the vector space modeling method to transform discrete knowledge change events into continuous directional adjustments. Compared with the rule-based empirical adjustment, this method realizes the computability and interpretability of the evolution of preference components, ensuring that the recommendation strategy keeps up with the latest medical progress in a timely manner and avoiding the instability of preference drift caused by knowledge updates. In the technical implementation process, multi-dimensional impact factor evaluation and vector space synthesis algorithms are integrated, providing a reliable mathematical modeling framework for dealing with complex knowledge evolution.

[0080] To construct a personalized recommendation system that can dynamically adapt to the evolution of medical knowledge, cognitive labels are generated by aligning the time of real-time behavior with the stability analysis of historical behavior and combining the knowledge hierarchy structure. The innovation lies in establishing a dynamic mapping between behavioral data and knowledge versions, quantifying long-term cognitive precipitation using cumulative distribution, and reflecting the professional system structure through hierarchical superposition, so as to realize the transformation from discrete behavior to structured cognitive level. In some embodiments, in step 102, in the dynamic knowledge graph, the obtained user real-time diagnosis and treatment behavior data and user historical medical interaction records are multi-dimensionally fused with the medical concepts to generate a set of user portrait labels with the user's medical cognitive level as the stratification benchmark, including: 501. Extract the operation timestamp and operation type identifier from the user's real-time diagnosis and treatment behavior data, as well as the behavior frequency and behavior duration from the user's historical medical interaction records; In step 501, the operation timestamp refers to the specific time information when the user performs diagnosis and treatment behaviors (such as consulting literature, prescribing medications). The operation type identifier is a classification code for distinguishing the types of diagnosis and treatment behaviors (such as diagnosis category A01, medication category B02). The behavior frequency represents the statistical value of the number of times the user accesses a specific medical concept. The behavior duration refers to the time difference between the start and end times of a single interaction behavior.

[0081] In the embodiments of the present application, first, the user's real-time diagnosis and treatment behavior data is collected as a log through the medical information system interface, and the operation timestamp (such as "2023-08-20 14:30:00") and operation type identifier are extracted using regular expressions. Then, the user's historical medical interaction records within three months are retrieved from the historical database, and the behavior frequency of each medical concept is calculated using the sliding window statistical method (such as the average monthly access to "hypertension" is 8 times), and the behavior duration is calculated through the timestamp difference (such as the duration of a single literature reading is 25 minutes).

[0082] 502. Align the operation timestamp with the time attribute of the medical concept in the dynamic knowledge graph in time sequence, and perform behavior matching according to the operation type identifier and the category attribute of the medical concept to obtain a preliminary association set; In step 502, the time sequence alignment is to match the time when the user behavior occurs with the effective time period of the medical concept in the knowledge graph. The behavior matching is to calculate the semantic similarity to associate the operation type with the field to which the medical concept belongs. The category attribute refers to the characteristics or labels used to describe and distinguish different medical concepts in the dynamic knowledge graph. These attributes help to clarify the specific classification to which each medical concept belongs. The preliminary association set is an intermediate data set containing the mapping relationship between user behavior and medical concepts.

[0083] In the embodiments of the present application, first, the dynamic time warping algorithm is used to perform temporal alignment between the user operation timestamps and the time attributes of medical concepts in the knowledge graph. Then, a medical behavior ontology library is constructed, and the cosine similarity is used to calculate the matching degree between the operation type identifier and the medical concept category attribute (such as "drug treatment plan"), and the behavior-concept pairs with a similarity threshold exceeding 0.7 are screened. Finally, the temporal alignment and behavior matching results are integrated to form a preliminary association set containing time validity markers.

[0084] 503. Calculate the stability weight associated with the medical concept in the user's historical medical interaction records according to the cumulative distribution ratio of the behavior frequency and behavior duration; In step 503, the behavior duration refers to the length of time spent by the user when performing a specific activity or operation. In the field of medical and health, it can refer to the specific duration of the user's participation in a certain health interaction. The cumulative distribution ratio is the distribution concentration of the user's historical behaviors in the time dimension. The stability weight reflects the persistence and regularity of the user's attention to specific medical concepts.

[0085] In the embodiments of the present application, first, the user's historical medical interaction records within three years are processed by time slicing, and the coefficient of variation of the behavior frequency and behavior duration of each medical concept in each time slice (such as a quarter) is calculated. Then, the exponential smoothing method is used to assign higher weights to the recent behavior data, and the cumulative distribution ratio of the behavior distribution is calculated through the Gini coefficient. Finally, the frequency stability (accounting for 60% of the weight) and duration stability (accounting for 40% of the weight) are linearly combined to generate a stability weight in the range of 0-1.

[0086] 504. Hierarchically superimpose the medical concepts in the preliminary association set with the stability weight, and combine the hierarchical distribution ratio of the medical concepts in the dynamic knowledge graph to generate a user portrait label set with the user's medical cognitive level as the stratification benchmark.

[0087] In step 504, hierarchical superposition is to perform weighted fusion of the preliminary association relationship and the stability weight according to the hierarchical structure of the medical concepts. The hierarchical distribution ratio refers to the hierarchical weight of the medical concepts in the knowledge graph in the disciplinary system (such as the basic medicine layer accounting for 30% and the clinical medicine layer accounting for 70%).

[0088] In the embodiments of the present application, first, a medical concept hierarchical tree is constructed, and the level where each node is located is determined according to the upper and lower relationships of the medical concepts in the dynamic knowledge graph (such as "coronary heart disease" belonging to the third-level clinical concept). Then, each medical concept in the preliminary association set is multiplied by its stability weight, and then combined with the hierarchical distribution ratio of the medical concepts in the dynamic knowledge graph. Finally, the hierarchical clustering algorithm is used to map the weighted concept association degree to the primary, intermediate, and advanced medical cognitive level labels, forming a user portrait label set containing the stratification benchmark with weight values.

[0089] The following is a specific example: In the application scenario of the cardiovascular department of a certain tertiary hospital, the user views the "Guidelines for Heart Failure of a Certain Edition" in real time (the operation type is guideline access). The historical record shows that the user has continuously paid attention to "Natriuretic Peptide Detection" in the past two years (with an average of 12 visits per month) and studies it for an average of 38 minutes each time. In step 501, the operation timestamp and the guideline access identifier of the user are extracted, and the behavioral frequency stability weight of 0.85 for "Natriuretic Peptide Detection" is calculated. In step 502, the guideline access time is aligned with the effective period of the heart failure treatment guideline in the knowledge graph, and the concept of "Natriuretic Peptide Monitoring" is matched. In step 503, the stability weight of 0.76 for the concept of "Natriuretic Peptide Detection" is calculated. In step 504, "Natriuretic Peptide Monitoring" is superimposed with the stability weight, and combined with its distribution ratio of 0.6 in the diagnostic standard layer, a hierarchical label set including "High-level Cognitive of Cardiac Markers" (weight 0.68) is finally generated.

[0090] To sum up, steps 501 to 504 innovatively solve the problems of static and single-dimensional user portraits in the medical field by deeply integrating real-time behavioral data with the spatio-temporal characteristics of the knowledge graph. The system can dynamically capture the spatio-temporal correlation between users' diagnosis and treatment behaviors and the evolution of medical knowledge, and construct a multi-level cognitive portrait with time sensitivity and disciplinary structure characteristics by combining historical behavior stability analysis and concept hierarchy weight assignment. Compared with traditional methods, this solution significantly improves the professionalism and timeliness of the user label system, accurately reflects the user's immediate knowledge needs, and deeply depicts the professional cognitive structure formed in the long term, providing a precise cognitive level reference benchmark for personalized medical services. The spatio-temporal alignment algorithm and hierarchical fusion mechanism adopted in the technical implementation provide a reliable computing framework for processing complex medical behavior data.

[0091] To accurately associate user behaviors with the dynamically evolving medical knowledge system, ensure the validity of concepts through time window comparison, and guarantee category relevance through semantic mapping. This method innovatively combines dynamic programming algorithms with deep semantic matching to solve the problems of timing misalignment and semantic gap in the association between behaviors and concepts. By generating association pairs with intensity values, it provides a reliable data basis for spatio-temporal two-dimensional verification in subsequent analysis. In some embodiments, in step 502, the operation timestamp is temporally aligned with the time attribute of the medical concept in the dynamic knowledge graph, and behavior matching is performed according to the operation type identifier and the category attribute of the medical concept to obtain a preliminary association set, including: 601. Perform a time window overlapping comparison between the operation timestamp and the time attribute of the medical concept in the dynamic knowledge graph. If the operation timestamp is within the effective time range of the time attribute, it is determined that the user's real-time diagnosis and treatment behavior is temporally aligned with the medical concept; In step 601, the time window overlapping comparison refers to verifying the intersection of the time period when the user behavior occurs and the valid time period of the knowledge graph concept. The effective time range is the start and end time periods during which the medical concept in the knowledge graph is recognized by the academic community. The operation timestamp is used to mark the exact time when the user performs certain medical-related activities (such as viewing a specific medical popular science article, receiving a certain treatment, etc.). The user's real-time diagnosis and treatment behavior describes the activities or behaviors that the user is currently performing related to medical care, including seeing a doctor, receiving a diagnosis or treatment, consulting health information, etc.

[0092] In the embodiment of the present application, first, the time attribute metadata of the target medical concept is extracted from the dynamic knowledge graph version library, including the concept effective start time (such as January 1st of a certain year) and the end time (marked as permanently valid if not expired). Then, the dynamic programming algorithm is used to perform a window overlapping comparison between the operation timestamp (such as 09:30 on March 15th of a certain year) and the concept effective start time, and a time series alignment is generated when the conditions "operation time ≥ effective start time" and "operation time ≤ effective end time" are met.

[0093] 602. According to the preset mapping relationship between the operation type identifier and the category attribute of the medical concept, match the category association between the user's real-time diagnosis and treatment behavior and the medical concept; In step 602, the preset mapping relationship is a classification correspondence rule library established based on the medical behavior ontology, including the semantic association matrix between the operation type (such as laboratory report interpretation) and the concept category (such as test index analysis). The category association refers to the matching relationship established between the specific behavior type of the user (such as click, read, share, etc.) and the category attribute of a specific medical concept in the dynamic knowledge graph.

[0094] In the embodiment of the present application, first, a mapping knowledge base of medical behaviors and concept categories is constructed, and the preset mapping relationship between the operation type identifier and the category attribute of the medical concept (such as "imaging examination appointment" corresponding to "imaging diagnosis standard") is screened in the mapping knowledge base by means of expert annotation. Then, a semantic similarity calculation model is introduced, and the text semantic matching degrees of the operation type identifier, the concept category, and the preset mapping relationship (such as "laboratory test index") are analyzed through the bidirectional encoder representation technology, and the matching pairs with a similarity threshold exceeding 0.75 are screened. Finally, a category association with a confidence value is generated.

[0095] 603. Combine the user's real-time diagnosis and treatment behavior with a time series alignment and a matching category association with the medical concept into a behavior-concept association pair, summarize each behavior-concept association pair, and obtain a preliminary association set.

[0096] In step 603, the behavior-concept association pair is a valid matching unit verified by both time and space, including the proof of timestamp alignment, category matching evidence, and association strength value. The preliminary association set is a data set composed of the real-time medical treatment behaviors of users and the corresponding medical concepts after temporal alignment and category association matching.

[0097] In the embodiments of the present application, first, the Cartesian product operation is performed on the temporal alignment result of step 601 and the category matching result of step 602, and the combinations that simultaneously meet the time validity and category matching are retained. Then, the behavior-concept association pair is calculated based on the confidence of the mapping relationship (such as 0.85) and the time window overlap degree (such as 1.0 for complete inclusion and 0.6 for partial overlap). Finally, all eligible behavior-concept association pairs are summarized and sorted in descending order of association strength to form a preliminary association set with weight identifiers.

[0098] The following is a specific example: In the hypertension management scenario of a certain tertiary hospital, the user performed the operation of "Interpretation of Ambulatory Blood Pressure Monitoring Report" on May 20th of a certain year (timestamp 05.20 14:00), and the effective time of the concept of "24-hour ambulatory blood pressure diagnostic criteria" in the knowledge graph was from November to December of a certain year. Step 601 confirmed through time window comparison that the operation time was within the effective range. Step 602 matched the operation type of "Report Interpretation" with the concept category of "Diagnostic Criteria" according to the preset mapping (similarity 0.89). Step 603 generated an association pair ("Interpretation of Ambulatory Blood Pressure Monitoring Report", "24-hour ambulatory blood pressure diagnostic criteria", association strength 0.93), which together with other valid associations formed a preliminary association set.

[0099] To sum up, steps 601 to 603 effectively solve the spatio-temporal misalignment problem between medical behaviors and knowledge concepts through a dual-verification mechanism. The system innovatively combines time validity verification and semantic association matching, which not only ensures that the knowledge concepts associated with user behaviors are in the current valid state but also guarantees the semantic consistency between the operation intention and the concept category. Compared with the single-dimensional matching method, this solution significantly improves the timeliness accuracy and professional relevance of the association results, avoiding the recommendation of outdated knowledge or semantically deviated content. The dynamic programming time comparison and semantic depth matching algorithms adopted in the technical implementation provide a reliable computational framework for processing complex medical spatio-temporal data, supporting the construction of a knowledge service system that accurately reflects the real-time needs of users.

[0100] To dynamically adapt to the evolution of medical knowledge and the changes in users' cognition, this solution calculates the vector similarity to match the core requirements, sets hierarchical adjustment rules to balance key coverage and knowledge expansion, integrates the dual factors of time decay and hot-spot enhancement, and adopts a diversity control strategy to break through the homogenization limitation of traditional recommendations. This method realizes the collaborative optimization of personalized demand matching, knowledge timeliness maintenance, and content ecological balance. In some embodiments, in step 104, according to the joint matching result between the dynamic recommendation feature vector and the medical concepts in the dynamic knowledge graph, the recommendation priority ranking of medical popular science articles is dynamically adjusted to generate a personalized recommendation sequence that matches the user's real-time medical cognitive level, including: 701. Calculate the matching degree between the association relationship of the dynamic recommendation feature vector and the medical concepts in the dynamic knowledge graph, and extract the matching degree value directly associated with the medical concept in the dynamic recommendation feature vector; In step 701, the matching degree calculation is to evaluate the association tightness between the dynamic recommendation feature vector and the medical concept in the knowledge graph through the vector space similarity measurement method. The matching degree value is the normalized similarity score, ranging from 0 to 1.

[0101] In the embodiments of the present application, first, the graph embedding technology is used to convert the medical concepts in the dynamic knowledge graph into vector representations to form a concept vector library. Then, the cosine similarity algorithm is used to calculate the similarity score of the association relationship between the user's dynamic recommendation feature vector (such as [0.34, -0.12, 0.78]) and the medical concepts in the dynamic knowledge graph. Finally, through the maximum-minimum normalization process, the similarity score is mapped to the 0-1 interval, and the matching degree value of the direct association exceeding the threshold of 0.6 is extracted.

[0102] 702. Set the priority adjustment rules for different medical concept levels according to the range of the matching degree value; In step 702, the priority adjustment rule is a weight distribution strategy divided according to the matching degree interval, including three-layer mechanisms: the core concept strengthening rule, the associated concept expansion rule, and the marginal concept suppression rule. The medical concept level is a way to organize rich medical information (such as diseases, symptoms, treatment methods, drugs, etc.), and these concepts are divided into different levels according to factors such as their nature, relevance, or importance.

[0103] In the embodiments of the present application, first, a three - level interval of the matching degree value is set. A high matching degree (0.8 - 1.0) corresponds to the core concept layer, a medium matching degree (0.6 - 0.8) corresponds to the associated concept layer, and a low matching degree (<0.6) corresponds to the marginal concept layer. Then, a hierarchical weight coefficient matrix is constructed through the medical concept hierarchy. The core concept layer is given 3 times the basic weight, the associated concept layer is given 1.5 times the basic weight, and the marginal concept layer is given 0.3 times the attenuated weight. Finally, a priority adjustment rule including parameters such as weight coefficients and upper limits of recommended frequencies is generated according to the three - level interval of the matching degree value.

[0104] 703. Based on the priority adjustment rule, perform priority weight assignment on the medical popular science articles associated with the user's real - time medical cognitive level in the dynamic knowledge graph; In step 703, the priority weight assignment is a comprehensive calculation process combining the concept - level weight and the article correlation degree, including two dynamic parameters: a time decay factor and a hot - spot strengthening factor. A medical popular science article refers to an article that popularizes medical knowledge, usually covering contents such as disease prevention, diagnosis, and treatment methods, with the goal of improving the public's health awareness.

[0105] In the embodiments of the present application, first, extract the associated concepts and their matching degree values of each medical popular science article in the dynamic knowledge graph, and convert them into hierarchical weights according to the rules in step 702. Then, introduce a time decay function to reduce the weight of articles that exceed the validity period (such as the weight of articles published more than 2 years is multiplied by 0.7), and superimpose the real - time popularity coefficient provided by the hot - spot monitoring module (such as the weight of new articles within 3 months is multiplied by 1.2). Finally, calculate the comprehensive priority weight assignment of the article through a linear weighting formula, retaining two decimal places of precision.

[0106] 704. According to the priority weight assignment result, screen medical popular science articles that match the user's real - time medical cognitive level from the dynamic knowledge graph, and perform a recommended priority ranking from high to low according to the priority weight assignment ratio to generate a personalized recommendation sequence.

[0107] In step 704, the recommended priority ranking is based on the descending order result of the weight values, and an adaptive queue management mechanism is used to dynamically maintain the recommendation sequence. The user's real - time medical cognitive level refers to the user's current understanding of medical knowledge, interest points, and the mastery of specific medical concepts at the current moment. This cognitive level is dynamically evaluated based on the user's real - time diagnosis and treatment behavior data (such as recently received treatments, consulted health information, etc.) and historical medical interaction records. The personalized recommendation sequence refers to an ordered list of a series of recommended contents customized according to the user's specific needs and preferences, specifically referring to the medical popular science articles that match the user's real - time medical cognitive level screened from the dynamic knowledge graph according to the priority weight assignment result.

[0108] In the embodiments of the present application, first, a candidate article pool is established based on the priority weight allocation results, and the article entries of medical popular science articles with a weight value lower than 0.5 are filtered. Then, the heap sort algorithm is used to sort the shortlisted articles according to the weight value for the recommended priority, generating an initial recommendation sequence. Finally, a diversity control module is introduced to implement an interval arrangement strategy for articles of the same category in the initial recommendation sequence (such as articles of the same disease type are spaced more than 3 positions apart), forming a final personalized recommendation sequence.

[0109] The following is a specific example: In the education scenario for heart failure patients in a certain tertiary hospital, the matching degree between the user's dynamic recommendation feature vector and the concept of "natriuretic peptide monitoring" is 0.92 (core layer), and the matching degree with the concept of "diuretic use" is 0.75 (associated layer). In step 701, the matching degree values of these two concepts are extracted; in step 702, the core layer is given a weight of 3 times and the associated layer is given a weight of 1.5 times; in step 703, the weight of a certain journal (associated with the core layer, published 3 months ago) is calculated, and the obtained weight value is 2.76, and the weight of another journal (associated layer, published 18 months ago) is 0.9; in step 704, after sorting, articles related to natriuretic peptide are recommended first, and "Diet Management for Heart Failure" is inserted in the third place to ensure diversity.

[0110] In summary, steps 701 to 704 effectively solve the dual problems of insufficient personalization and lack of knowledge timeliness in the medical recommendation system through a multi-level weight dynamic allocation mechanism. The system innovatively fuses and calculates the user's real-time cognitive characteristics, knowledge evolution status, and content quality factors, achieving three key breakthroughs. First, a quantitative mapping model between concept matching degree and content priority is established to accurately reflect the user's cognitive needs. Second, a dual-factor dynamic adjustment of time decay and hotspot strengthening is introduced to balance the accuracy and frontier of knowledge. Third, a diversity control strategy is adopted to avoid the homogenization of recommendation results. Compared with traditional static recommendation methods, this solution significantly improves the dynamic fit between the recommended content and the user's cognitive state, providing an intelligent knowledge service adaptation ability for scenarios such as patient education and clinical decision support.

[0111] Figure 2 The following is a schematic structural diagram of a medical popular science article recommendation system based on a user portrait provided by the embodiments of the present application, as Figure 2 shown, the system includes: An analysis module 21, which constructs a dynamic knowledge graph containing the dynamic dependency relationship between medical concepts by analyzing the logical association between unstructured text and structured case data in multi-source heterogeneous medical data; A fusion module 22, in the dynamic knowledge graph, fuses the obtained user's real-time diagnosis and treatment behavior data and the user's historical medical interaction records with the medical concepts in multiple dimensions to generate a set of user portrait tags based on the user's medical cognitive level as the stratification benchmark; A conversion module 23 that performs cross-scenario feature conversion on the user's medical preference features in the user portrait tag set from the unrealistic scenarios in the user's historical medical interaction records to the realistic scenarios of the user's real-time diagnosis and treatment behavior data, generating a dynamic recommendation feature vector that is updated synchronously with the evolution trend of medical knowledge; An adjustment module 24 that dynamically adjusts the recommended priority ranking of medical popular science articles according to the joint matching result between the dynamic recommendation feature vector and the medical concepts in the dynamic knowledge graph, generating a personalized recommendation sequence that matches the user's real-time medical cognitive level.

[0112] Figure 2 The described medical popular science article recommendation system based on user portraits can execute Figure 1 The described medical popular science article recommendation method based on user portraits in the illustrated embodiments, the implementation principle and technical effects will not be elaborated further. For the medical popular science article recommendation system based on user portraits in the above embodiments, the specific ways in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0113] In a possible design, Figure 2 The medical popular science article recommendation system based on user portraits in the illustrated embodiments can be implemented as a computing device, such as Figure 3 As shown, the computing device can include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0114] The processing component 32 is used for the above Figure 1 The medical popular science article recommendation method based on user portraits in the described embodiments.

[0115] Among them, the processing component 32 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component can also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0116] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0117] Of course, the computing device may also necessarily include other components, such as an input / output interface, a display component, a communication component, and the like.

[0118] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above-mentioned peripheral interface module can be an output device, an input device, and the like.

[0119] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0120] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device can refer to a cloud server, and the above-mentioned processing component, storage component, etc. can be basic server resources leased or purchased from a cloud computing platform.

[0121] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above Figure 1 shown embodiment of a method for recommending medical popular science articles based on user portraits.

[0122] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0123] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0124] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for recommending medical popular science articles based on user portraits, characterized in that, including: constructing a dynamic knowledge graph containing the dynamic dependence relationship between medical concepts by analyzing the logical association between unstructured text and structured case data in multi-source heterogeneous medical data; in the dynamic knowledge graph, fusing the obtained user real-time diagnosis and treatment behavior data and user historical medical interaction records with the medical concepts in multiple dimensions to generate a set of user portrait labels with the user's medical cognitive level as the stratification benchmark; performing cross-scenario feature transformation on the user medical preference features in the set of user portrait labels from the unrealistic scenarios of the user historical medical interaction records to the realistic scenarios of the user real-time diagnosis and treatment behavior data to generate a dynamic recommendation feature vector that is updated synchronously with the evolution trend of medical knowledge; dynamically adjusting the recommended priority ranking of medical popular science articles according to the joint matching result between the dynamic recommendation feature vector and the medical concepts in the dynamic knowledge graph to generate a personalized recommendation sequence that matches the user's real-time medical cognitive level.

2. The method according to claim 1, characterized in that, The performing cross-scenario feature transformation on the user medical preference features in the set of user portrait labels from the unrealistic scenarios of the user historical medical interaction records to the realistic scenarios of the user real-time diagnosis and treatment behavior data to generate a dynamic recommendation feature vector that is updated synchronously with the evolution trend of medical knowledge includes: extracting the preference components of the unrealistic scenarios of the user historical medical interaction records according to the user medical preference features in the set of user portrait labels; constructing a non-linear projection relationship from the unrealistic scenarios to the realistic scenarios by analyzing the implicit association pattern between the operation trajectories of medical concepts in the unrealistic scenarios and the user real-time diagnosis and treatment behavior data in the realistic scenarios; based on the non-linear projection relationship, mapping the user medical preference features to the feature space of the realistic scenarios through the preference components of the unrealistic scenarios, and separating the scenario noise components irrelevant to the user's medical cognitive level in the feature space, and retaining the core preference components that remain stable and transmitted in the cross-scenario feature transformation; by tracking the change of the update status of the medical concepts, calculating the dynamic offset of the core preference components in the evolution trend of medical knowledge, and superimposing the dynamic offset on the core preference components to generate a dynamic recommendation feature vector containing the evolution trend of medical knowledge.

3. The method according to claim 2, characterized in that, The by tracking the change of the update status of the medical concepts, calculating the dynamic offset of the core preference components in the evolution trend of medical knowledge includes: obtaining the change record of the association relationship of the medical concepts in the dynamic knowledge graph, and extracting the newly added dependence relationship and the invalidated dependence relationship between medical concepts to obtain the update status mark of the evolution trend of medical knowledge; based on the update status mark, determining the positive offset direction corresponding to the newly added dependence relationship and the negative offset direction corresponding to the invalidated dependence relationship in the core preference components; obtaining the dynamic offset of the core preference components in the evolution trend of medical knowledge according to the superposition effect of the positive offset direction and the negative offset direction, and combining the proportional weights of the newly added dependence relationship and the invalidated dependence relationship.

4. The method according to claim 3, characterized in that, Determining the positive offset direction corresponding to the newly added dependency relationship and the negative offset direction corresponding to the invalidated dependency relationship in the core preference component based on the update status flag includes: Based on the update status flag, mapping the newly added dependency relationship to a first adjustment parameter of the core preference component, and mapping the invalidated dependency relationship to a second adjustment parameter of the core preference component; Generating a first set of direction vectors according to the product of the association strength between medical concepts connected by each newly added dependency relationship and the first adjustment parameter, and generating a second set of direction vectors according to the product of the historical association strength between medical concepts corresponding to each invalidated dependency relationship and the second adjustment parameter; Superposing the direction vectors in the first set of direction vectors to obtain the positive offset direction, and superposing the direction vectors in the second set of direction vectors to obtain the negative offset direction.

5. The method according to claim 1, characterized in that, In the dynamic knowledge graph, performing multi-dimensional fusion on the obtained user real-time diagnosis and treatment behavior data, user historical medical interaction records and the medical concepts to generate a set of user portrait labels with the user's medical cognitive level as the stratification benchmark, including: Extracting the operation timestamp and operation type identifier in the user real-time diagnosis and treatment behavior data, and the behavior frequency and behavior duration in the user historical medical interaction records; Performing time series alignment on the operation timestamp and the time attribute of the medical concept in the dynamic knowledge graph, and performing behavior matching according to the operation type identifier and the category attribute of the medical concept to obtain a preliminary association set; Calculating the stability weight of the association between medical concepts in the user historical medical interaction records according to the cumulative distribution ratio of the behavior frequency and behavior duration; Performing hierarchical superposition of the medical concepts in the preliminary association set and the stability weight, and combining the hierarchical distribution ratio of the medical concepts in the dynamic knowledge graph to generate a set of user portrait labels with the user's medical cognitive level as the stratification benchmark.

6. The method according to claim 5, characterized in that, The performing time series alignment on the operation timestamp and the time attribute of the medical concept in the dynamic knowledge graph, and performing behavior matching according to the operation type identifier and the category attribute of the medical concept to obtain a preliminary association set includes: Performing time window overlapping comparison on the operation timestamp and the time attribute of the medical concept in the dynamic knowledge graph. If the operation timestamp is within the effective time range of the time attribute, it is determined that the user real-time diagnosis and treatment behavior is time series aligned with the medical concept; Matching the category association of the user real-time diagnosis and treatment behavior and the medical concept according to the preset mapping relationship between the operation type identifier and the category attribute of the medical concept; Combining the user real-time diagnosis and treatment behavior and medical concept with time series alignment and matching category association into a behavior and concept association pair, and summarizing each behavior and concept association pair to obtain a preliminary association set.

7. The method according to claim 1, characterized in that, Dynamically adjusting the recommended priority ranking of medical popular science articles according to the joint matching result between the dynamic recommended feature vector and the medical concepts in the dynamic knowledge graph, and generating a personalized recommendation sequence that matches the user's real-time medical cognitive level, including: Calculating the matching degree of the association relationship between the dynamic recommended feature vector and the medical concepts in the dynamic knowledge graph, and extracting the matching degree values directly associated with the medical concepts in the dynamic recommended feature vector; Setting priority adjustment rules for different medical concept levels according to the range of the matching degree values; Based on the priority adjustment rules, performing priority weight allocation on the medical popular science articles associated with the user's real-time medical cognitive level in the dynamic knowledge graph; According to the priority weight allocation result, screening medical popular science articles that match the user's real-time medical cognitive level from the dynamic knowledge graph, and performing recommended priority ranking from high to low according to the priority weight allocation ratio to generate a personalized recommendation sequence.

8. A medical popular science article recommendation system based on user portraits, characterized in that, Including: An analysis module that constructs a dynamic knowledge graph containing dynamic dependence relationships between medical concepts by analyzing the logical association between unstructured text and structured case data in multi-source heterogeneous medical data; A fusion module that multi-dimensionally fuses the obtained user's real-time diagnosis and treatment behavior data and the user's historical medical interaction records with the medical concepts in the dynamic knowledge graph to generate a set of user portrait labels with the user's medical cognitive level as the stratification benchmark; A conversion module that performs cross-scenario feature conversion on the user's medical preference features in the set of user portrait labels from the non-real scenario of the user's historical medical interaction records to the real scenario of the user's real-time diagnosis and treatment behavior data, and generates a dynamic recommended feature vector that is updated synchronously with the evolution trend of medical knowledge; An adjustment module that dynamically adjusts the recommended priority ranking of medical popular science articles according to the joint matching result between the dynamic recommended feature vector and the medical concepts in the dynamic knowledge graph, and generates a personalized recommendation sequence that matches the user's real-time medical cognitive level.

9. A computing device, characterized in that, Including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for recommending medical popular science articles based on user portraits as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements a method for recommending medical popular science articles based on user portraits as described in any one of claims 1 to 7.

Images