Medical question and answer management method based on machine learning and related equipment

Abstract

The invention provides a medical question and answer management method based on machine learning and related equipment, and is applied to the technical field of data processing. The method comprises the following steps: acquiring medical education application scene information, and a target large language model and target user information matched with the medical education application scene information; preprocessing the medical document information to generate text feature classification information; processing the medical teaching question and answer data to generate initial medical entity feature associated information; processing the text feature classification information and the initial medical entity feature association information to generate node information and node association information matched with the node information; processing the node information and the node association information to generate medical knowledge graph information; acquiring real-time department information of the target hospital; and processing the real-time department information, the medical knowledge graph information and the target user information of the target hospital based on the target large language model to generate question and answer information of the target user.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a medical question and answer management method based on machine learning and related equipment. Background Art

[0002] In the prior art, patients usually rely on doctors or medical professionals to obtain information, which limits the accessibility and coverage of information. Patients need to make appointments and go to medical institutions, which takes time and energy, and can only get help at specific time points. In some areas or specific fields, medical resources may be scarce, making it difficult for patients to obtain medical information and advice. There is information asymmetry between doctors and patients. Doctors have professional knowledge, while patients may have limited knowledge of the medical field, resulting in difficulties in communication and understanding. In the traditional consultation mode, patients may have to wait for a long time to get a response or appointment from the doctor. After consulting a doctor, patients may have deviations or omissions in their understanding of the doctor's advice due to differences in memory and understanding. Some patients may be reluctant to face the doctor directly for consultation because of subjective questions that are difficult to talk about or embarrassment. Traditional consultations may be accompanied by higher medical costs, especially in some specific healthcare systems.

[0003] In recent years, the types of Chinese medical knowledge graphs have evolved from general medical knowledge graphs to single-disease medical knowledge graphs, and the types of diseases involved have become increasingly diverse, and good results have been achieved in terms of application. However, the existing Chinese single-disease medical knowledge graphs are mostly for more common diseases, and how to use knowledge graphs to assist in the diagnosis and treatment of rare diseases is also an issue that needs to be solved in the future.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of this application is to provide a medical question and answer management method and related equipment based on machine learning, which at least overcomes the problems existing in the prior art to a certain extent, by collecting medical education scene information, documents and question and answer data, preprocessing medical documents to generate text features, and extracting medical entity association information from question and answer data. Combine this information to build a medical knowledge graph, characterize decision information, and obtain real-time department information of the target hospital to match decision execution. Use training samples and historical clinical guideline data to train a large language model to generate a target model, which combines real-time department and knowledge graph information to generate personalized question and answer information for users. The accuracy of the model is ensured by calculating the frequency of abnormal co-occurrence, generating adjacent features, and determining the sampling ratio. Ultimately, the model integrates medical knowledge, real-time data, and user needs to provide accurate and personalized medical consultation and improve the quality of medical services.

[0006] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0007] According to one aspect of the present application, a medical question and answer management method based on machine learning is provided, comprising: obtaining medical education application scenario information, a target large language model and target user information that match the medical education application scenario information, wherein the medical education application scenario information includes medical document information and medical teaching question and answer data; preprocessing the medical document information to generate text feature classification information, wherein the text feature classification information respectively represents attribute information of different medical entity features; processing the medical teaching question and answer data to generate initial medical entity feature association information, wherein the initial medical entity feature association information is used to characterize the association between different medical entity features; and performing preprocessing on the text feature classification information and the initial medical entity feature. The target hospital comprises a medical knowledge graph, a medical education application scenario, a medical knowledge graph, a medical information graph, a medical knowledge graph information and a medical education application scenario information. The target hospital comprises a medical knowledge graph, a medical knowledge graph information and a medical information graph. The target hospital comprises a medical knowledge graph, a medical information ...

[0008] In one embodiment of the present application, obtaining a target large language model that matches the medical education application scenario information includes: obtaining a training sample set and a preset large language model that matches the medical education application scenario information; obtaining a historical clinical guideline document data set; processing the historical clinical guideline document data set to obtain abnormal node features and abnormal co-occurrence frequencies that match the abnormal node features; processing the abnormal node features and the abnormal co-occurrence frequencies that match the abnormal node features to generate a number of node information; processing the node information to generate an original feature library; classifying the original feature library to generate user feature information, wherein the user The feature information is used to characterize the physiological characteristics of the same user for different diseases; the user feature information is processed to generate a feature mapping table, wherein the feature mapping table is used to characterize the different physiological characteristics of different patients for the same disease; the original feature library is processed based on the feature mapping table to generate target feature data, wherein the target feature data is used to characterize the node information and decision path information of the clinical guideline document information processing; the preset large language model is trained based on the training data set with the target feature data to generate a target large language model; the method also includes a calculation formula for calculating the abnormal co-occurrence frequency, and the calculation formula is: in, is the probability of A and B co-occurring, are the probabilities of A and B appearing alone, respectively.

[0009] In one embodiment of the present application, the preset large language model is trained based on the training data set with the target feature data to generate the target large language model, including: obtaining any number of data features in the training data set with the target feature data; generating adjacent features based on the distance between any number of data features and other numbers of data features of the same category, wherein the adjacent features include a preset number of the data features of the any number; determining a sampling ratio based on the number of each data feature in the training set; determining a sampling ratio based on the sampling ratio; sampling the adjacent features based on the sampling ratio to generate a preset number of sampling features; and generating a preset number of sampling features based on the distance between any data feature and each sampling Features, generate multiple groups of data groups, wherein each group of data groups includes a preset number of data samples, wherein each group of data groups includes a preset number of data samples, wherein at least one data sample includes identification information; process the multiple groups of data groups to generate a training set and a verification set, wherein the training set and the verification set both include a number of data features; train the preset large language model based on the training set to generate a trained large language model; process the trained large language model based on the verification set to generate a verification result; if the data sample containing the identification information in the verification result is node information and decision path information, then use the trained large language model as the target large language model.

[0010] In one embodiment of the present application, the text feature classification information and the initial medical entity feature association information are processed to generate node information and node association information matching the node information, including: processing the medical document information based on the initial medical entity feature association information to generate node information; processing the text feature classification information to generate clinical symptom classification information; processing the clinical symptom classification information to generate symptom attribute information of different users, wherein the symptom attribute information of different users is used to characterize the correlation between different users for different symptoms; processing the symptom attribute information of different users to generate node correlation information, wherein the node correlation information is used to characterize the correlation between different nodes.

[0011] In one embodiment of the present application, the medical document information is processed based on the initial medical entity feature association information to generate node information, including: processing the medical document information based on the initial medical entity feature association information to generate correlation information of medical entity features, wherein the correlation information of medical entity features is used to characterize the correlation between different medical entity features; processing the correlation information of the medical entity to generate node information.

[0012] In one embodiment of the present application, the node information and the node association information are processed to generate medical knowledge graph information, including: processing the initial node information based on the node association information to generate a decision adjustment factor; processing the initial node information and the decision information matching the initial node information based on the decision adjustment factor to generate preset node information and preset decision information matching the preset node information; generating medical knowledge graph information based on the preset node information and the preset decision information.

[0013] In one embodiment of the present application, the real-time department information of the target hospital, the medical knowledge graph information and the target user information are processed based on a target large language model to generate question and answer information for the target user, including: processing the target user information and the medical knowledge graph information based on a target large language model to generate initial consultation information for the target user; processing the real-time department information of the target hospital based on the initial consultation information of the target user to generate target department research and development information; processing the target department research and development information based on a target large language model to generate question and answer information for the target user, wherein the question and answer information for the target user is used to represent the consulting suggestions of the target hospital to the target user.

[0014] Another aspect of the present application is a medical question and answer management device based on machine learning, characterized in that it includes: an acquisition module for acquiring medical education application scenario information, a target large language model and target user information matching the medical education application scenario information, wherein the medical education application scenario information includes medical document information and medical teaching question and answer data; acquiring real-time department information of a target hospital, wherein the real-time department information of the target hospital includes real-time scientific research information of each department and real-time personnel information of each department, and the real-time department information of the target hospital is used to characterize decision execution information matching the medical knowledge graph information; a processing module for preprocessing the medical document information to generate text feature classification information, wherein the text feature classification information respectively represents The method comprises the steps of: processing the text feature classification information and the initial medical entity feature association information to generate node information and node association information matching the node information; processing the node information and the node association information to generate medical knowledge graph information, wherein the medical knowledge graph information is used to characterize decision information matching the medical education application scenario information; processing the real-time department information of the target hospital, the medical knowledge graph information and the target user information based on the target large language model to generate question and answer information for the target user.

[0015] According to another aspect of the present application, an electronic device is characterized in that it includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the above-mentioned machine learning-based medical question and answer management method by executing the executable instructions.

[0016] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a second processor, the above-mentioned medical question and answer management method based on machine learning is implemented.

[0017] According to another aspect of the present application, a computer program product is provided, comprising a computer program, wherein the computer program implements the above-mentioned medical question and answer management method based on machine learning when executed by a third processor.

[0018] The present application provides a machine learning-based medical question-and-answer management method and related equipment, in which the server collects medical education scene information, medical documents, and medical teaching question-and-answer data, and pre-processes the medical documents to generate text feature classification information, extracts initial medical entity feature association information from the medical teaching question-and-answer data, and characterizes the association between different medical entities. Combined with the text feature classification information and the medical entity feature association information, node information and node association information are generated to construct a medical knowledge graph, characterize decision information, and obtain real-time department information of the target hospital, including scientific research information and personnel information, to match the decision execution information in the medical knowledge graph.

[0019] Using the training sample set and the historical clinical guideline document data set, the preset large language model is trained to generate the target large language model. Based on the target large language model, the real-time department information and medical knowledge graph information are combined to generate the target user's question and answer information, and provide personalized medical consultation suggestions. The accuracy and effectiveness of the model are ensured by calculating the abnormal co-occurrence frequency, generating adjacent features, determining the sampling ratio, generating data groups, training and verifying the model. The target large language model is used to process user information and medical knowledge graph information to generate initial consultation information, and then combined with real-time department information to generate the final question and answer information. By integrating medical knowledge, real-time data and user needs, the large language model is used to generate accurate and personalized medical question and answer information, which improves the quality and efficiency of medical services.

[0020] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1A flowchart of a medical question and answer management method based on machine learning provided by an embodiment of the present application is shown; Figure 2 A schematic diagram of the structure of a medical question and answer management device based on machine learning provided in one embodiment of the present application is shown; Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present application is shown; Figure 4 A schematic diagram of a storage medium provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0022] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0023] Combine the following Figure 1 To describe the medical question and answer management method based on machine learning according to the exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown to facilitate understanding of the spirit and principles of the present application, and the embodiments of the present application are not limited in this regard. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0024] In one embodiment, the present application also proposes a medical question and answer management method and related equipment based on machine learning. Figure 1 The following schematically shows a flow chart of a medical question and answer management method based on machine learning according to an embodiment of the present application. Figure 1 As shown, the method is applied to a server, comprising: S101, obtaining medical education application scenario information, a target large language model matching the medical education application scenario information, and target user information.

[0025] In one embodiment, the medical education application scenario information includes medical document information and medical teaching question and answer data. For medical document information, this type of data includes: medical textbooks, which are standard textbooks covering basic medicine, clinical medicine, pharmacy, nursing and other fields. Clinical guidelines, which are guiding documents on disease diagnosis, intervention and management issued by professional medical organizations and institutions. Expert consensus, which are recommendations and practice standards on specific medical issues jointly formulated by experts in the field. Scientific literature, including journal articles, conference papers, research reports, etc., which provide the latest medical research results and discoveries.

[0026] For medical teaching question and answer data, this type of data includes: question and answer records of students and doctors, questions and answers collected in actual teaching and clinical practice. Mock exams and quizzes, exam question banks for medical education, including questions and standard answers. Online forums and question and answer platforms, such as discussion posts in medical professional forums and question and answer websites. Data processing is performed on the above medical document information and medical teaching question and answer data, including text cleaning, formatting, deduplication and error correction, semantic annotation of text, and identification of key medical concepts and entities. The processed data is integrated into the knowledge base to support knowledge retrieval technologies such as RAG, ensuring that the collected data is both comprehensive and representative, thereby providing a solid foundation for the development of an efficient and accurate medical education question and answer system.

[0027] In order to obtain the target large language model that matches the medical education application scenario, a series of detailed steps are required to process and prepare data, as well as train and optimize the model. The following is a detailed description of this process: obtain a training sample set, a preset large language model that matches the medical education application scenario information, obtain a training sample set, collect question-answer pairs from medical education platforms, forums, clinical cases, and textbooks, and have medical experts annotate the question-answer data to ensure the accuracy and reliability of the data. Select a pre-trained large language model suitable for processing natural language, such as GPT, BERT, etc., and fine-tune the model to better adapt to the specific needs of medical education.

[0028] Obtain a historical clinical guideline document dataset, obtain historical clinical guideline documents from medical institutions and public health organizations, and convert the documents into a processable format, such as a text file. Process the historical clinical guideline document dataset to obtain abnormal node features and abnormal co-occurrence frequencies that match the abnormal node features. Use natural language processing technology to extract keywords, concepts, and relationships in the document, and identify abnormal node features in the document, such as uncommon symptoms or intervention methods. Process the abnormal node features and the abnormal co-occurrence frequencies that match the abnormal node features to generate several node information, analyze the co-occurrence frequencies of the abnormal node features to identify important medical associations, and generate node information from the co-occurrence analysis, which will be used for subsequent feature library construction. The specific calculation method is detailed in the following text and will not be discussed in detail here.

[0029] Process the node information to generate the original feature library, classify the original feature library, and generate user feature information, where the user feature information is used to characterize the physiological characteristics of the same user for different diseases. For example, some people may feel much better after taking 1-2 medicines, while others may suffer from recurrence of other diseases due to physical factors. Integrate the node information into a unified feature library, and extract useful features from the text data, such as symptoms, diseases, and intervention methods. Analyze the physiological characteristics of users (such as patients) for different diseases, use the user feature information to characterize the user's reaction and intervention response to different diseases, process the user feature information, and generate a feature mapping table, where the feature mapping table is used to characterize the different physiological characteristics of different patients for the same disease. Analyze the different physiological characteristics of different patients for the same disease, and construct a feature mapping table for comparing and analyzing the characteristics of different patients. Process the original feature library based on the feature mapping table, extract node information related to clinical decision-making, and analyze and integrate decision path information to support clinical decision-making.

[0030] The original feature library is processed based on the feature mapping table to generate target feature data, wherein the target feature data is used to characterize the node information and decision path information of the clinical guideline document information processing, and the preset large language model is trained based on the training data set with the target feature data to generate the target large language model. Specifically, the patient features in the feature mapping table are matched with the data in the original feature library to ensure that each feature can be correctly mapped to the corresponding disease type and patient group, and inconsistent or inaccurate data is removed to ensure the quality of the data used for training. According to the feature mapping table, the most useful features for clinical decision-making, such as symptom severity, complication risk, etc., are selected, and the selected features are integrated and optimized to form a target feature data set, which will be used to train the model to identify decision nodes and paths in clinical guidelines.

[0031] The target feature data should be able to clearly describe each key node in the clinical guidelines, such as diagnostic criteria, intervention options, and preventive measures. Similarly, the target feature data needs to be able to reflect the entire decision path from diagnosis to intervention, including possible branches and decision points. When the target feature data is ready, use this data to train the preset large language model to generate a target large language model specifically for medical education and clinical applications: format and preprocess the target feature data to meet the input requirements of the model, select a suitable pre-trained large language model, such as BERT, GPT, etc., as the base model for fine-tuning. Set training parameters such as learning rate, batch size, training cycle, etc. If there is enough labeled data, supervised learning methods can be used for training, and the performance of the model can be evaluated through the validation set and test set to ensure the accuracy and reliability of the model. According to the validation results, the model is tuned and the model parameters and structure are optimized. Ensure that the generated large language model can not only understand and process complex medical texts, but also provide effective support and decision-making assistance in actual medical education and clinical environments.

[0032] In another implementation, the present application also includes a calculation formula for calculating the abnormal co-occurrence frequency, and the specific calculation formula is: in, is the probability of A and B co-occurring, and are the probabilities of A and B appearing alone, respectively. The purpose of this formula is to identify and quantify whether there is a significant co-occurrence relationship between different features (such as symptoms, test results, etc.) in clinical guideline documents. This co-occurrence relationship may indicate the importance and relevance of these features in specific disease interventions. By calculating PMI, it can help extract meaningful information from large amounts of medical data, improve the accuracy of interventions and interventions, analyze clinical text data, and discover potential connections between diseases, symptoms, and interventions. This information can be used to build intervention decision trees and optimize the clinical decision-making process.

[0033] In another embodiment, a preset large language model is trained based on a training data set with target feature data to generate a target large language model, including: obtaining any number of data features in the training data set with target feature data, generating adjacent features based on the distance between any number of data features and other numbers of data features of the same category, wherein the adjacent features include a preset number of any number of data features, determining a sampling ratio based on the number of each data feature in the training set, determining a sampling ratio based on the sampling ratio, sampling adjacent features based on the sampling ratio, and generating a preset number of sampled features.

[0034] Based on any data feature and each sampling feature, multiple data groups are generated, wherein each data group includes a preset number of data samples, wherein each data group includes a preset number of data samples, wherein at least one data sample includes identification information, multiple data groups are processed to generate training sets and verification sets, wherein both the training sets and the verification sets include a number of data features, and a preset large language model is trained based on the training set to generate a trained large language model. The trained large language model is processed based on the verification set to generate a verification result, and if the data sample containing the identification information in the verification result is node information and decision path information, the trained large language model is used as the target large language model.

[0035] Specifically, any number of data features are obtained from the training data set, and based on the distance between any number of data features and other numbers of data features of the same category, features adjacent to any number of data features are generated according to a preset number. The number of each data feature in the training set is obtained, a sampling ratio is determined according to the number of features, a sampling ratio is determined based on the sampling ratio, adjacent features are sampled based on the sampling ratio, and a preset number of sampling features are generated. Based on any data feature and each sampling feature, multiple data groups are generated, each data group contains a preset number of data samples, wherein at least one data sample includes identification information, and the multiple data groups are processed to generate a training set and a validation set, ensuring that both the training set and the validation set include a number of data features.

[0036] The preset large language model is trained based on the training set to generate a trained large language model, and the trained large language model is processed based on the validation set to generate a validation result. Check whether the data sample containing identification information in the validation result is node information and decision path information. If the validation result meets the conditions, the trained large language model is used as the target large language model. This process involves multiple links such as data preprocessing, feature engineering, model training and validation, ensuring that the model can accurately learn and predict key information in medical education and clinical decision-making. In this way, the performance and applicability of the model can be improved, so that it can better serve medical education and clinical practice.

[0037] S102: pre-process the medical document information to generate text feature classification information.

[0038] In one embodiment, preprocessing medical document information and generating text feature classification information is an important step in the application of natural language processing (NLP) in the medical field. This process involves extracting useful information from raw medical documents and converting it into structured data that can be understood by machine learning models. Specifically, medical documents such as clinical guidelines, research papers, case reports, etc. are collected, irrelevant content such as headers, footers, advertisements, etc. are removed, and the text is normalized, such as unifying fonts and formats. The text in the document is broken down into separate words or terms, and the part of speech of each word is marked (noun, verb, adjective, etc.). Medical entities in the text, such as diseases, drugs, symptoms, test results, etc., are identified, and the identified entities are mapped to medical ontologies or dictionaries to ensure the accuracy and consistency of the entities.

[0039] Identify key phrases in the text, such as "hypertension management", "diabetes intervention", etc., and extract contextual information related to specific entities, such as drug dosage, intervention duration, etc. Classify the extracted features into different attribute categories, such as cause, symptoms, diagnostic methods, intervention plans, etc., and identify the relationship between entities, such as the relationship of "drug-intervention-disease". Construct a feature vector for each text, containing the entity features and attribute information to which it belongs, and represent the feature vector as the attribute information of different medical entity features.

[0040] The extracted entities and relationships are constructed into a knowledge graph for further analysis and application, and the processed data is stored in a structured database, such as a relational database or a graph database. The processed data is used to train machine learning models, such as classification models and prediction models, and the performance of the models is verified through experiments to ensure the accuracy of feature extraction and classification. Through these steps, the original medical documents can be converted into structured information that can be used for machine learning and data analysis, thereby supporting applications such as medical research, clinical decision support systems, and personalized medicine, which not only improves the efficiency of data processing, but also enhances the interpretability and reliability of the model.

[0041] S103: Process the medical teaching question and answer data to generate initial medical entity feature association information.

[0042] In one implementation, processing medical teaching question and answer data to generate initial medical entity feature association information and using a graph database to store this information is a key step in building a medical knowledge graph. Specifically, preprocessing the medical teaching question and answer data includes removing irrelevant content, such as punctuation marks, special characters, etc., breaking down the text into meaningful words or terms, and using named entity recognition (NER) technology to identify medical entities in the text, such as diseases, symptoms, and drugs. Extracting attributes related to medical entities from the text, such as common symptoms of diseases, side effects of drugs, etc., identifies the relationship between medical entities in the text, such as the association between "diabetes" and "hyperglycemia."

[0043] Combine the extracted entities and attributes into the initial medical entity feature association information, and use the association information to characterize the association between different medical entity features. Map the extracted medical entities to the corresponding concepts in the authoritative biomedical ontology database, and verify and enrich the association relationship between entities according to the medical ontology. Select a suitable graph database system, such as Neo4j, ArangoDB, etc., create nodes in the graph database to represent medical entities, and create edges in the graph database to represent the association relationship between medical entities. Use the graph database to build a medical entity association graph, and create indexes for the nodes and edges in the graph to optimize retrieval performance.

[0044] Using the retrieval capability of the graph database, we query the information in the graph according to the needs, extract the medical entities and relationships related to the query, and use them for subsequent applications. We regularly update the medical entities and relationships in the graph database to reflect the latest medical knowledge, maintain the graph database system, and ensure its performance and security. In this way, medical teaching question and answer data can be converted into a structured medical knowledge graph, providing strong data support for medical education and clinical practice. This graph can not only provide fast and accurate medical information retrieval, but also support complex data analysis and decision making.

[0045] S104, processing the text feature classification information and the initial medical entity feature association information to generate node information and node association information matching the node information.

[0046] In one implementation, the medical document information is processed based on the initial medical entity feature association information to generate node information, the text feature classification information is processed to generate clinical symptom classification information, and the clinical symptom classification information is processed to generate symptom attribute information of different users, wherein the symptom attribute information of different users is used to characterize the association of different users with different symptoms. The symptom attribute information of different users is processed to generate node association information, wherein the node association information is used to characterize the association between different nodes.

[0047] Specifically, we analyze the entity feature association information in medical documents, such as the relationship between diseases and symptoms, and between diseases and intervention methods, and represent medical entities (such as diseases, symptoms, and drugs) as nodes in a graph database. We classify the text features extracted from medical documents, paying special attention to features related to clinical symptoms, and classify symptoms according to their attributes (such as physiological, psychological, pathological, etc.). We further analyze the symptom data of different users (such as different patient groups), and establish a symptom attribute profile for each user, recording their symptom characteristics and associations, such as the severity and duration of the symptoms.

[0048] Analyze the user's symptom attribute information, identify the commonalities and differences between different users, and represent the correlation between different nodes (symptoms, diseases, interventions, etc.) in the graph database, such as co-occurrence frequency, causal relationship, etc. Use the generated node information and node correlation information to build a medical knowledge graph to ensure the accuracy and completeness of the graph, and continuously optimize the graph through expert verification and data update. As new medical research and clinical data are generated, the knowledge graph is regularly updated, and the graph database and related software systems are maintained to ensure their performance and security. Medical document information can be effectively converted into a structured medical knowledge graph to provide support for medical education and clinical practice. This graph can help users better understand the complex relationships between medical concepts and provide data support for medical research and clinical decision-making.

[0049] In another embodiment, the medical document information is processed based on the initial medical entity feature association information to generate node information, including: processing the medical document information based on the initial medical entity feature association information to generate association information of medical entity features, wherein the association information of medical entity features is used to characterize the association between different medical entity features. Use natural language processing (NLP) technology, such as named entity recognition (NER), to identify medical entities such as diseases, symptoms, drugs, intervention methods, etc. from medical documents, and extract relevant features for each identified medical entity, which may include symptom descriptions, drug dosages, intervention results, etc. Analyze the association between different medical entity features, such as the association between a symptom and a specific disease, or the relationship between a drug and an intervention effect, and represent this association information as structured data, such as a relationship triple (entity 1, relationship, entity 2), for subsequent processing and analysis.

[0050] The correlation information of medical entities is processed to generate node information. Medical entities and their features are created as nodes in the graph database. Each node represents a medical entity or concept. According to the correlation information of medical entity features, edges are created between corresponding nodes to represent the relationship between them.

[0051] By creating nodes and edges, a complete medical knowledge graph is constructed. This graph reflects the complex relationship between medical entities and their characteristics. The graph is optimized to ensure the accuracy and completeness of the information, which may include adding more relevance information, correcting erroneous associations, and enhancing the queryability of the graph. Through this process, the information in medical documents is converted into a dynamic knowledge graph that can be used in various applications, improving the availability and practicality of medical information.

[0052] S105, processing the node information and the node association information to generate medical knowledge graph information.

[0053] In one embodiment, the initial node information is processed based on the node correlation information to generate a decision adjustment factor. Based on the correlation information of medical entity features, the connection and interaction between different medical concepts are analyzed, and key factors are extracted from the node correlation information. These factors can be used as a basis for adjusting clinical decisions, such as drug interactions, risk of disease complications, etc.

[0054] Based on the decision adjustment factor, the initial node information and the decision information matching the initial node information are processed to generate the preset node information and the preset decision information matching the preset node information. The decision adjustment factor is applied to the initial node information to adjust and optimize the relevant medical decisions, and the preset node information is generated based on the adjusted decision information, which reflects the optimized medical concepts and decisions. The preset node information is matched with the corresponding decision information to ensure that each node has corresponding decision support, and detailed decision information is generated for each preset node, such as intervention recommendations, preventive measures, etc.

[0055] Generate medical knowledge graph information based on preset node information and preset decision information. Medical knowledge graph information is used to represent decision information that matches the medical education application scenario information. Integrate preset node information and preset decision information into the medical knowledge graph to ensure that the medical knowledge graph can clearly represent decision information that matches the medical education application scenario. In clinical practice, use the medical knowledge graph to assist doctors in making diagnosis and intervention decisions. As new medical research and clinical data are generated, regularly update the nodes and decision information in the medical knowledge graph, maintain the graph database and related software systems to ensure their performance and security, evaluate the effectiveness and accuracy of the medical knowledge graph information through actual application cases, and verify whether the decision information in the graph is consistent with the latest medical research and clinical practice.

[0056] Through this process, the medical knowledge graph can not only provide information about the associations between medical concepts, but also support the adjustment and optimization of medical education and clinical decision-making. This dynamic and adjustable knowledge graph is a valuable resource for medical education and clinical practice.

[0057] S106, obtaining real-time department information of the target hospital.

[0058] In one implementation, the real-time department information of the target hospital includes real-time scientific research information of each department and real-time personnel information of each department, and the real-time department information of the target hospital is used to characterize the decision execution information that matches the medical knowledge graph information. Obtaining the real-time department information of the target hospital is an important step in constructing decision execution information that matches the medical knowledge graph. The following are the detailed steps to achieve this goal: First, determine the information needs and clarify what scientific research related information needs to be collected, such as ongoing scientific research projects, published scientific research papers, transformation of scientific research results, etc., and determine the type of personnel information that needs to be collected, such as the doctor's professional field, work experience, duty arrangements, etc.

[0059] Negotiate with the target hospital to establish a data interface to ensure that real-time department information can be obtained safely and legally, and that the method of obtaining data complies with data protection regulations and the hospital's privacy policy. Use API or other data capture tools to regularly obtain real-time department information from the target hospital's information system, and ensure that the acquired data can be updated in real time to reflect the latest situation of the department. Clean the acquired data, remove invalid or inaccurate information, and integrate data from different sources into a unified data set for easy analysis and application. Extract real-time scientific research information of each department from the processed data, and extract real-time personnel information of each department from the processed data.

[0060] Match real-time department information with decision information in the medical knowledge graph to characterize the actual situation of decision execution, and update the medical knowledge graph based on real-time department information to ensure that the graph reflects the latest decision execution information. Use real-time department information to support clinical decision-making, such as personnel scheduling, resource allocation, etc., and use real-time scientific research information for scientific research project management and results tracking. Continuously monitor the quality and integrity of data to ensure the accuracy of information, regularly check and maintain data interfaces with target hospitals, and ensure continuous data acquisition. The real-time department information of the target hospital can be effectively acquired and utilized to support the real-time update and application of decision execution information in the medical knowledge graph, which helps to improve the quality and efficiency of medical services and promote the development of medical research.

[0061] S107, based on the target large language model, the real-time department information of the target hospital, the medical knowledge graph information and the target user information are processed to generate question and answer information for the target user.

[0062] In one implementation, the target user information and medical knowledge graph information are processed based on the target large language model to generate the target user's initial consultation information, and the target hospital's real-time department information is processed based on the target user's initial consultation information to generate the target department's research and development information. The target department's research and development information is processed based on the target large language model to generate the target user's question and answer information, wherein the target user's question and answer information is used to represent the target hospital's consulting suggestions to the target user.

[0063] This process describes how to use the target large language model to process the target hospital's real-time department information, medical knowledge graph information, and target user information to generate personalized question-and-answer information for different users. The following are the detailed steps to achieve this goal: process the target user information and medical knowledge graph information, analyze the target user's health status, medical history, intervention needs and other information, combine the user information and the medical knowledge graph to generate the user's initial consultation information, which may include questions that the user may be concerned about or medical information that needs to be known. Match the user's initial consultation information with the target hospital's real-time department information to determine which departments may provide relevant services or expertise, and based on the matching results, generate department research and development information related to the user's consultation, such as the latest research results and intervention methods of a specific department.

[0064] Use the target large language model to process the target department's research and development information, generate personalized question and answer information for the target user, and ensure that the generated question and answer information is accurate, easy to understand, and meets the user's specific situation and needs. Integrate the generated question and answer information into the target hospital's consulting recommendations for the target user, verify the accuracy and applicability of the consulting recommendations, ensure that they meet medical standards and regulatory requirements, and provide the generated question and answer information to the user as a response to the user's consultation. Based on user feedback and further questions, continuously optimize the question and answer information and consulting recommendations. Monitor the impact of question and answer information on user satisfaction and health outcomes, and regularly evaluate the performance of the large language model to ensure the quality of the question and answer information it generates.

[0065] Through these steps, it can be ensured that the question-answer information generated by the target large language model is both accurate and personalized, meeting the medical consultation needs of different users while improving the quality and efficiency of medical services. This method can be widely used in fields such as online medical consultation, telemedicine, and personal health management.

[0066] In this application, the server collects medical education scene information, medical documents, and medical teaching question and answer data, and pre-processes the medical documents to generate text feature classification information, extracts initial medical entity feature association information from the medical teaching question and answer data, and characterizes the association between different medical entities. Combined with the text feature classification information and the medical entity feature association information, node information and node association information are generated to construct a medical knowledge graph, characterize decision information, and obtain real-time department information of the target hospital, including scientific research information and personnel information, to match the decision execution information in the medical knowledge graph.

[0067] Using the training sample set and the historical clinical guideline document data set, the preset large language model is trained to generate the target large language model. Based on the target large language model, the real-time department information and medical knowledge graph information are combined to generate the target user's question and answer information, and provide personalized medical consultation suggestions. The accuracy and effectiveness of the model are ensured by calculating the abnormal co-occurrence frequency, generating adjacent features, determining the sampling ratio, generating data groups, training and verifying the model. The target large language model is used to process user information and medical knowledge graph information to generate initial consultation information, and then combined with real-time department information to generate the final question and answer information. By integrating medical knowledge, real-time data and user needs, the large language model is used to generate accurate and personalized medical question and answer information, which improves the quality and efficiency of medical services.

[0068] In one embodiment, if Figure 2 As shown, the present application also provides a medical question and answer management device based on machine learning, comprising: Acquisition module 201 is used to acquire medical education application scenario information, a target large language model and target user information that match the medical education application scenario information, wherein the medical education application scenario information includes medical document information and medical teaching question and answer data; acquire real-time department information of a target hospital, wherein the real-time department information of the target hospital includes real-time scientific research information of each department and real-time personnel information of each department, and the real-time department information of the target hospital is used to represent decision execution information that matches the medical knowledge graph information; The processing module 202 is used to pre-process the medical document information to generate text feature classification information, wherein the text feature classification information respectively represents the attribute information of different medical entity features; process the medical teaching question and answer data to generate initial medical entity feature association information, wherein the initial medical entity feature association information is used to characterize the correlation between different medical entity features; process the text feature classification information and the initial medical entity feature association information to generate node information and node association information matching the node information; process the node information and the node association information to generate medical knowledge graph information, wherein the medical knowledge graph information is used to characterize decision information matching the medical education application scenario information; based on the target large language model, the real-time department information of the target hospital, the medical knowledge graph information and the target user information are processed to generate question and answer information for the target user.

[0069] The present application embodiment provides an electronic device, such as Figure 3 As shown, the electronic device 3 includes a first processor 300, a memory 301, a bus 302 and a communication interface 303. The first processor 300, the communication interface 303 and the memory 301 are connected via the bus 302; the memory 301 stores a computer program that can be run on the first processor 300, and when the first processor 300 runs the computer program, it executes the medical question and answer management method based on machine learning provided in any of the aforementioned embodiments of the present application.

[0070] The memory 301 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 303 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0071] The bus 302 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 301 is used to store programs, and the first processor 300 executes the programs after receiving the execution instruction. The medical question and answer management method based on machine learning disclosed in any implementation of the embodiments of the present application may be applied to the first processor 300, or implemented by the first processor 300.

[0072] The first processor 300 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the first processor 300. The above-mentioned first processor 300 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a readily available programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be embodied as a hardware decoding processor to execute, or a combination of hardware and software modules in the decoding processor to execute. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 301, and the first processor 300 reads the information in the memory 301 and completes the steps of the above method in combination with its hardware.

[0073] The electronic device provided in the above-mentioned embodiments of the present application and the medical question and answer management method based on machine learning provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0074] The present application embodiment provides a computer-readable storage medium, such as Figure 4 As shown, the computer-readable storage medium 401 stores a computer program, and when the computer program is read and executed by the second processor 402, the medical question and answer management method based on machine learning as described above is implemented.

[0075] The technical solution of the embodiment of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling an electronic device (which may be an air conditioner, a refrigeration device, a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the method described in the embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0076] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the medical question and answer management method based on machine learning provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0077] An embodiment of the present application provides a computer program product, including a computer program, wherein the computer program is executed by a third processor to implement the method described above.

[0078] The computer program product provided by the above-mentioned embodiments of the present application and the medical question and answer management method based on machine learning provided by the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0079] It should be noted that, in this application, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or still includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0080] Each embodiment in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the evaluation of the medical question and answer management method based on machine learning, the electronic device, the electronic device, and the readable storage medium embodiment are basically similar to the above-mentioned embodiment of the medical question and answer management method based on machine learning, so the description is relatively simple, and the relevant parts can be referred to the partial description of the embodiment of the medical question and answer management method based on machine learning.

[0081] Although the present application is disclosed as above, the present application is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application shall be subject to the scope defined by the claims.

Claims

1. A medical question and answer management method based on machine learning, characterized in that: include: Acquire medical education application scenario information, a target large language model matching the medical education application scenario information, and target user information, wherein the medical education application scenario information includes medical document information and medical teaching question and answer data; Preprocessing the medical document information to generate text feature classification information, wherein the text feature classification information respectively represents attribute information of different medical entity features; Processing the medical teaching question and answer data to generate initial medical entity feature association information, wherein the initial medical entity feature association information is used to characterize the association between different medical entity features; Processing the text feature classification information and the initial medical entity feature association information to generate node information and node association information matching the node information; Processing the node information and the node association information to generate medical knowledge graph information, wherein the medical knowledge graph information is used to represent decision information that matches the medical education application scenario information; Acquire the real-time department information of the target hospital, wherein the real-time department information of the target hospital includes the real-time scientific research information of each department and the real-time personnel information of each department, and the real-time department information of the target hospital is used to represent the decision execution information matching the medical knowledge graph information; The real-time department information of the target hospital, the medical knowledge graph information and the target user information are processed based on the target large language model to generate question and answer information for the target user.

2. The method according to claim 1, characterized in that Obtain a target large language model that matches the medical education application scenario information, including: Obtain training sample sets and preset large language models that match medical education application scenario information; Obtain historical clinical guideline document datasets; Processing the historical clinical guideline document dataset to obtain abnormal node features and abnormal co-occurrence frequencies matching the abnormal node features; Processing the abnormal node features and the abnormal co-occurrence frequencies matching the abnormal node features to generate a number of node information; Process the node information to generate the original feature library; Classifying the original feature library to generate user feature information, wherein the user feature information is used to characterize the physiological features of the same user for different diseases; Processing the user characteristic information to generate a characteristic mapping table, wherein the characteristic mapping table is used to characterize different physiological characteristics of different patients for the same disease; Processing the original feature library based on the feature mapping table to generate target feature data, wherein the target feature data is used to characterize node information and decision path information of clinical guideline document information processing; Training the preset large language model based on the training data set with target feature data to generate a target large language model; The method also includes a calculation formula for calculating the abnormal co-occurrence frequency, and the calculation formula is: ； in, is the probability of A and B co-occurring, and are the probabilities of A and B appearing alone, respectively.

3. The method according to claim 2, characterized in that The preset large language model is trained based on the training data set with target feature data to generate a target large language model, including: Get any number of data features in the training data set with target feature data; Generate adjacent features based on the distance between any number of data features and other number of data features of the same category, wherein the adjacent features include a preset number of the any number of data features; Determining a sampling ratio based on the number of each data feature in the training set; Based on the sampling ratio, determining a sampling ratio; Sampling the adjacent features based on the sampling ratio to generate a preset number of sampling features; Based on any data feature and each sampling feature, generating a plurality of data groups, wherein each data group comprises a preset number of data samples, wherein each data group comprises a preset number of data samples, wherein at least one data sample comprises identification information; Processing multiple data sets to generate training sets and validation sets, wherein both the training sets and the validation sets include a number of data features; Training the preset large language model based on the training set to generate a trained large language model; Processing the trained large language model based on the verification set to generate a verification result; If the data sample containing identification information in the verification result is node information and decision path information, the trained large language model is used as the target large language model.

4. The method according to claim 1, characterized in that Processing the text feature classification information and the initial medical entity feature association information to generate node information and node association information matching the node information includes: Processing the medical document information based on the initial medical entity feature association information to generate node information; Processing the text feature classification information to generate clinical symptom classification information; Processing the clinical symptom classification information to generate symptom attribute information of different users, wherein the symptom attribute information of different users is used to characterize the relevance of different symptoms of different users; The symptom attribute information of the different users is processed to generate node correlation information, wherein the node correlation information is used to characterize the correlation between different nodes.

5. The method according to claim 4, characterized in that The medical document information is processed based on the initial medical entity feature association information to generate node information, including: Processing the medical document information based on the initial medical entity feature association information to generate medical entity feature association information, wherein the medical entity feature association information is used to characterize the association between different medical entity features; The association information of the medical entity is processed to generate node information.

6. The method according to claim 4, characterized in that Processing the node information and the node association information to generate medical knowledge graph information includes: Processing the initial node information based on the node relevance information to generate a decision adjustment factor; Processing the initial node information and the decision information matching the initial node information based on the decision adjustment factor to generate preset node information and preset decision information matching the preset node information; Generate medical knowledge graph information based on the preset node information and preset decision information.

7. The method according to claim 6, characterized in that The target hospital's real-time department information, the medical knowledge graph information, and the target user information are processed based on the target large language model to generate question-and-answer information for the target user, including: Processing the target user information and the medical knowledge graph information based on the target large language model to generate initial consultation information for the target user; Processing the real-time department information of the target hospital based on the initial consultation information of the target user to generate target department research and development information; The target department research and development information is processed based on the target large language model to generate question and answer information of the target user, wherein the question and answer information of the target user is used to represent the consulting suggestions of the target hospital to the target user.

8. A medical question and answer management device based on machine learning, characterized in that: The device comprises: An acquisition module is used to acquire medical education application scenario information, a target large language model and target user information that match the medical education application scenario information, wherein the medical education application scenario information includes medical document information and medical teaching question and answer data; acquire real-time department information of a target hospital, wherein the real-time department information of the target hospital includes real-time scientific research information of each department and real-time personnel information of each department, and the real-time department information of the target hospital is used to represent decision execution information that matches the medical knowledge graph information; A processing module is used to pre-process the medical document information to generate text feature classification information, wherein the text feature classification information respectively represents the attribute information of different medical entity features; process the medical teaching question and answer data to generate initial medical entity feature association information, wherein the initial medical entity feature association information is used to characterize the correlation between different medical entity features; process the text feature classification information and the initial medical entity feature association information to generate node information and node association information matching the node information; process the node information and the node association information to generate medical knowledge graph information, wherein the medical knowledge graph information is used to characterize decision information matching medical education application scenario information; based on the target large language model, the real-time department information of the target hospital, the medical knowledge graph information and the target user information are processed to generate question and answer information for the target user.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to execute the medical question and answer management method based on machine learning as described in any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the medical question and answer management method based on machine learning described in any one of claims 1 to 7 is implemented.

Images