Intelligent question answering system based on multi-source medical knowledge retrieval enhancement

Abstract

The invention provides an intelligent question answering system based on multi-source medical knowledge retrieval enhancement, and belongs to the technical field of medical knowledge retrieval. The system comprises a user interaction module used for obtaining medical questions; the problem analysis model is used for analyzing medical problems and identifying and extracting key information in the medical problems; the knowledge management module is used for constructing a medical knowledge base by maintaining and updating a knowledge graph, a medical term library and a medical literature database; carrying out vectorization processing on the medical literature; constructing a vector index so as to retrieve and access the medical literature; the medical knowledge retrieval module is used for retrieving related medical literature nodes according to the key information by using the vector index to obtain corresponding multi-source medical knowledge information; the answer generation module is used for integrating the multi-source medical knowledge information and generating multi-source medical knowledge fragments; and inputting the multi-source medical knowledge fragments into a pre-trained large language model to generate comprehensive summary answers.

Description

Technical Field

[0001] The present invention relates to the technical field of medical knowledge retrieval, and more particularly to an intelligent question-answering system based on multi-source medical knowledge retrieval enhancement. Background Art

[0002] Traditional question-answering systems mostly rely on keyword matching and template generation. These methods work well for simple queries, but they often have difficulty providing accurate and in-depth answers when faced with complex medical questions. In recent years, the development of knowledge graphs and large language model generation technologies has provided new possibilities for improving the performance of question-answering systems. Knowledge graphs can organize and represent complex medical concepts and relationships, while large language generation technologies can express this knowledge in the form of natural language. Some systems use retrieval-augmented generation (RAG) methods to enhance the training of generative models by retrieving large amounts of text data to improve the accuracy and relevance of generated answers. However, the application of these systems in the medical field is not mature enough, and the types of knowledge sources are relatively single. Summary of the invention

[0003] In response to the above problems, the purpose of the present invention is to provide an intelligent question-answering system based on multi-source medical knowledge retrieval enhancement, which can deeply integrate, reason and comprehensively judge multi-source medical knowledge from knowledge graphs, medical terminology bases and medical literature, and by applying retrieval-enhanced generation technology, it can extract information from multi-dimensional knowledge and generate accurate, comprehensive and highly explainable answers.

[0004] In order to achieve the above object, the present invention is implemented through the following technical solutions: The present invention discloses an intelligent question-answering system based on multi-source medical knowledge retrieval enhancement, comprising: A user interaction module, used for obtaining medical questions input by users and sending the medical questions to the question parsing model; A question parsing model is used to analyze medical questions using natural language processing technology, identify and extract key information in medical questions; the key information includes: keywords, phrases and query intent; The knowledge management module is used to build a medical knowledge base by maintaining and updating the knowledge graph, medical terminology base and medical literature database; vectorize the medical literature and convert it into a numerical vector form; and build a vector index to retrieve and access the medical literature; The medical knowledge retrieval module is used to use the vector index to retrieve relevant medical literature nodes and obtain corresponding multi-source medical knowledge information based on the key information extracted by the question analysis module; The answer generation module is used to integrate multi-source medical knowledge information and generate multi-source medical knowledge fragments; the multi-source medical knowledge fragments are input into a pre-trained large-scale language model, and the deep semantic understanding ability of the large model is used to generate a comprehensive summary answer.

[0005] Furthermore, the medical problems include: Medication consultation, drug Q&A, health Q&A, and disease-related epidemiological information.

[0006] Furthermore, the use of natural language processing technology to analyze medical problems and identify and extract key information in medical problems includes: The BiLSTM-CRF model is used to segment medical questions into words or phrases. A deep learning-based NER system is used to identify specific entities in medical questions; specific entities include: names, places, and organizations; Determine the user's query intention based on the LSTM network.

[0007] Furthermore, the medical knowledge base is constructed by maintaining and updating the knowledge graph, medical terminology database and medical literature database, including: Collecting medical text data, including clinical guidelines, industry standards, diagnosis and treatment specifications, and medical encyclopedias; Clean and standardize the collected medical text data, remove irrelevant information, and ensure the quality and consistency of the medical text data; Use named entity recognition technology to identify medical terms from medical text data and determine the corresponding medical entities; Mapping identified medical terms to a standardized medical terminology set; Extract the relationships between medical entities and standardize the attributes of medical entities and the relationships between medical entities; align medical entities and their ontologies through pattern matching to unify the conceptual systems in different data sources and generate a knowledge graph; Review and verify the content of the knowledge graph.

[0008] Furthermore, the construction of a medical knowledge base by maintaining and updating the knowledge graph, medical terminology database and medical literature database also includes: Construct the structure of medical terminology and determine the fields, content and attributes of medical terms; Import the corresponding information of medical terms into the medical term list structure, edit the medical term list, and generate a medical term library; Perform version management of medical terminology database.

[0009] Furthermore, the construction of a medical knowledge base by maintaining and updating the knowledge graph, medical terminology database and medical literature database also includes: Collect medical literature, journals, conference papers, clinical guidelines, and case reports as medical documents; Perform data preprocessing on medical documents, remove irrelevant information in medical documents, and format medical documents; Classify medical documents according to medical fields and document types; Based on different classifications, medical documents are divided into text segments according to size; Use the BERT model to vectorize text fragments and store the text vectors in a medical literature database.

[0010] Further, the constructing of the vector index includes: Store the text vector into the index structure; Set cosine similarity as the similarity metric for comparing text vectors.

[0011] Furthermore, the vector index is used to retrieve relevant medical literature nodes according to the key information extracted by the question analysis module to obtain corresponding multi-source medical knowledge information, including: Convert the key information extracted by the question parsing module into a query vector; According to the query vector, use the vector index to find the text vector similar to the query vector and calculate the corresponding cosine similarity; The found text vectors are sorted in descending order according to cosine similarity, and the relevant medical documents are returned as multi-source medical knowledge information.

[0012] Furthermore, the multi-source medical knowledge information is integrated to generate multi-source medical knowledge fragments, including: Based on the cosine similarity of the text vectors corresponding to the multi-source medical knowledge information, the top three text vectors ranked in descending order of cosine similarity are extracted to generate multi-source medical knowledge fragments.

[0013] Compared with the prior art, the beneficial effect of the present invention is that: the present invention discloses an intelligent question-answering system based on multi-source medical knowledge retrieval enhancement, which builds an in-depth and comprehensive medical knowledge base based on a large language model by combining medical knowledge graphs, professional terminology libraries and medical literature. This system uses document vectorization and efficient vector indexing technology to achieve rapid retrieval of massive medical data. When generating answers, it not only relies on a single knowledge source, but also comprehensively considers information from knowledge graphs, terminology libraries and literature libraries, and combines the deep semantic understanding ability of the large model to make comprehensive judgments and intelligent reasoning to generate accurate, in-depth and easy-to-understand medical answers.

[0014] It can be seen that compared with the prior art, the present invention has outstanding substantive features and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0016] Figure 1 It is a system structure diagram of a specific implementation mode of the present invention.

[0017] In the figure, 1. User interaction module; 2. Question parsing model; 3. Knowledge management module; 4. Medical knowledge retrieval module; 5. Answer generation module. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] See also Figure 1 As shown, this embodiment provides an intelligent question-answering system based on multi-source medical knowledge retrieval enhancement, including: a user interaction module 1, a question parsing model 2, a knowledge management module 3, a medical knowledge retrieval module 4 and an answer generation module 5.

[0020] The user interaction module 1 is used to obtain the medical questions input by the user and send the medical questions to the question parsing model.

[0021] In a specific implementation, the user inputs a medical question through the user interaction interface provided by the user interaction module 1, and passes it to the question parsing model 2. Among them, medical questions include medication consultation, drug questions and answers, health questions and answers, disease-related epidemiological information, etc., which are natural language inputs and are oriented to the general medical field.

[0022] The question parsing model 2 is used to analyze medical questions using natural language processing technology, identify and extract key information in medical questions; the key information includes: keywords, phrases and query intent.

[0023] In a specific implementation, the problem analysis model 2 is specifically used for: The BiLSTM-CRF model is used to segment medical questions into words or phrases. A deep learning-based NER system is used to identify specific entities in medical questions, including names, places, and organizations. The user's query intent is determined based on the LSTM network.

[0024] The analysis of medical problems is performed using natural language processing (NLP). As an example, the specific process includes the following: 1. Tokenization: Split the text input by the user into smaller units, such as words or phrases. For example, BiLSTM-CRF.

[0025] 2. Named Entity Recognition (NER): Identify specific entities in the text, such as names, places, organizations, etc. Entity recognition can use a deep learning-based NER system (such as BiLSTM-CRF).

[0026] 3. Intent Recognition: Determine the intent or purpose of the user input. Deep learning-based methods (such as LSTM) can be used.

[0027] The knowledge management module 3 is used to build a medical knowledge base by maintaining and updating the knowledge graph, medical terminology base and medical literature database; to vectorize the medical literature and convert it into a numerical vector form; and to build a vector index for retrieving and accessing medical literature.

[0028] In a specific implementation, the knowledge management module 3 is specifically used for: Build knowledge graphs, build medical terminology bases, build medical literature databases, and build vector indexes.

[0029] As an example, building a knowledge graph specifically includes the following process: (1) Data collection: The first step in building a medical knowledge graph is to collect large-scale medical text data, which may include clinical guidelines, industry standards, diagnosis and treatment specifications, medical encyclopedias, etc.

[0030] (2) Data preprocessing: Clean and standardize the collected data to remove irrelevant information and ensure data quality and consistency.

[0031] (3) Entity Recognition: Use natural language processing techniques, especially named entity recognition (NER), to identify medical concepts such as diseases, drugs, symptoms, etc. from text.

[0032] (4) Entity standardization: Map the identified medical terms to a standardized medical terminology set to ensure the consistency of the entities.

[0033] (5) Relationship extraction: Based on entity recognition, the relationship between entities is extracted, such as the relationship between disease and symptoms, the relationship between drugs and indications, etc.

[0034] (6) Knowledge alignment: Knowledge alignment is a meticulous process to ensure that medical information from different sources remains consistent and accurate in the knowledge graph. It first identifies and links the same medical entities, such as diseases or drugs, from different data sources, and then standardizes their attributes and relationships to eliminate information ambiguity and duplication. Next, the conceptual systems in different data sources are unified through pattern matching and ontology alignment.

[0035] (7) Knowledge verification: Medical experts review and verify the content of the knowledge graph to ensure the accuracy and authority of the information.

[0036] As an example, building a medical terminology database specifically includes the following processes: Construct the structure of medical terminology and determine the fields, content and attributes of medical terms.

[0037] Import the corresponding information of medical terms into the medical terminology table structure, edit the medical terminology table, and generate a medical terminology database.

[0038] Perform version management of medical terminology database.

[0039] As an example, building a medical literature database specifically includes the following processes: (1) Collect medical literature, journals, conference papers, clinical guidelines, and case reports as medical documents.

[0040] (2) Perform data preprocessing on medical documents, remove irrelevant information in medical documents, and format medical documents.

[0041] (3) Classify medical documents according to medical fields and document types.

[0042] (4) Based on different classifications, medical documents are divided into text segments according to size.

[0043] (5) Use the BERT model to vectorize text fragments and store the text vectors in the medical literature database.

[0044] As an example, constructing a vector index specifically includes the following process: Text preprocessing: Clean the original text data, including removing stop words and punctuation marks, performing stem extraction or word form restoration, etc.

[0045] Segmentation: Split text into individual words or phrases.

[0046] Vectorization: Convert text into vector form. You can use models such as bag-of-words model, TF-IDF vector, Word2Vec, BERT, etc. to generate vector representation of text.

[0047] Build the index: Store the text vectors into the index structure, which usually involves converting each document in the text collection into a vector and organizing it according to the requirements of the index structure.

[0048] Define similarity measure: Determine the similarity metric used to compare text vectors, such as cosine similarity.

[0049] Vector retrieval: When a query is received, the query text is converted into a vector and the index is used to quickly find the document vector that is most similar to the query vector.

[0050] Result sorting and returning: Based on the results of the similarity measurement, the retrieved documents are sorted and the most relevant documents are returned.

[0051] The medical knowledge retrieval module 4 is used to use the vector index to retrieve relevant medical literature nodes according to the key information extracted by the question analysis module, and obtain the corresponding multi-source medical knowledge information.

[0052] In a specific implementation, first, the key information extracted by the question parsing module is converted into a query vector; then, based on the query vector, a vector index is used to find a text vector similar to the query vector, and the corresponding cosine similarity is calculated; finally, the found text vectors are sorted in descending order according to the cosine similarity, and the relevant medical documents are returned as multi-source medical knowledge information.

[0053] The medical knowledge retrieval module 4 can use vector indexing technology to quickly retrieve relevant medical literature nodes based on the key information extracted by the question analysis module.

[0054] The answer generation module 5 is used to integrate multi-source medical knowledge information and generate multi-source medical knowledge fragments; the multi-source medical knowledge fragments are input into a pre-trained large-scale language model, and the deep semantic understanding ability of the large model is used to generate a comprehensive summary answer.

[0055] In a specific embodiment, the multi-source medical knowledge information is integrated to generate multi-source medical knowledge fragments, including: based on the cosine similarity of the text vectors corresponding to the multi-source medical knowledge information, the top three text vectors ranked in descending order of cosine similarity are extracted to generate multi-source medical knowledge fragments.

[0056] As an example, the answer generation module 5 first combines the retrieved multi-source medical knowledge to integrate the information, and then uses the generation capability of the large model, combined with comprehensive judgment and intelligent reasoning, to generate a comprehensive answer.

[0057] In summary, the present invention can deeply integrate, infer and comprehensively judge multi-source medical knowledge from knowledge graphs, medical terminology bases and medical literature. By applying retrieval-enhanced generation technology, it can extract information from multi-dimensional knowledge and generate accurate, comprehensive and highly explainable answers.

[0058] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. As for the method disclosed in the embodiment, since it corresponds to the system disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0059] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0060] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or units, which can be electrical, mechanical or other forms.

[0061] 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 distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0062] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, or each module may exist physically separately, or two or more modules may be integrated into one unit.

[0063] Similarly, each processing unit in each embodiment of the present invention may be integrated into one functional module, or each processing unit may exist physically, or two or more processing units may be integrated into one functional module.

[0064] Finally, it should be noted that, in this article, 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 also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0065] The above is a detailed introduction to the intelligent question-answering system based on multi-source medical knowledge retrieval enhancement provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the scope of protection of the present invention.

Claims

1. An intelligent question-answering system based on multi-source medical knowledge retrieval enhancement, characterized in that: include: A user interaction module, used for obtaining medical questions input by users and sending the medical questions to the question parsing model; Question parsing model, which is used to analyze medical questions using natural language processing technology, identify and extract key information from medical questions; The key information includes: keywords, phrases and query intent; The knowledge management module is used to build a medical knowledge base by maintaining and updating the knowledge graph, medical terminology base and medical literature database; vectorize the medical literature and convert it into a numerical vector form; and build a vector index to retrieve and access the medical literature; The medical knowledge retrieval module is used to use the vector index to retrieve relevant medical literature nodes and obtain corresponding multi-source medical knowledge information based on the key information extracted by the question analysis module; The answer generation module is used to integrate multi-source medical knowledge information and generate multi-source medical knowledge fragments; the multi-source medical knowledge fragments are input into a pre-trained large-scale language model, and the deep semantic understanding ability of the large model is used to generate a comprehensive summary answer.

2. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 1 is characterized in that: Medical problems include: Medication consultation, drug Q&A, health Q&A, and disease-related epidemiological information.

3. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 1 is characterized in that: The use of natural language processing technology to analyze medical problems and identify and extract key information in medical problems includes: The BiLSTM-CRF model is used to segment medical questions into words or phrases. A deep learning-based NER system is used to identify specific entities in medical questions; specific entities include: names, places, and organizations; Determine the user's query intention based on the LSTM network.

4. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 1 is characterized in that: The medical knowledge base is constructed by maintaining and updating the knowledge graph, medical terminology database and medical literature database, including: Collecting medical text data, including clinical guidelines, industry standards, diagnosis and treatment specifications, and medical encyclopedias; Clean and standardize the collected medical text data, remove irrelevant information, and ensure the quality and consistency of the medical text data; Use named entity recognition technology to identify medical terms from medical text data and determine the corresponding medical entities; Mapping identified medical terms to a standardized medical terminology set; Extract the relationships between medical entities and standardize the attributes of medical entities and the relationships between medical entities; align medical entities and their ontologies through pattern matching to unify the conceptual systems in different data sources and generate a knowledge graph; Review and verify the content of the knowledge graph.

5. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 1 is characterized in that: The medical knowledge base is constructed by maintaining and updating the knowledge graph, medical terminology database and medical literature database, and further includes: Construct the structure of medical terminology and determine the fields, content and attributes of medical terms; Import the corresponding information of medical terms into the medical term list structure, edit the medical term list, and generate a medical term library; Perform version management of medical terminology database.

6. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 1 is characterized in that: The medical knowledge base is constructed by maintaining and updating the knowledge graph, medical terminology database and medical literature database, and further includes: Collect medical literature, journals, conference papers, clinical guidelines, and case reports as medical documents; Perform data preprocessing on medical documents, remove irrelevant information in medical documents, and format medical documents; Classify medical documents according to medical fields and document types; Based on different classifications, medical documents are divided into text segments according to size; Use the BERT model to vectorize text fragments and store the text vectors in a medical literature database.

7. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 6 is characterized in that: The constructing of the vector index comprises: Store the text vector into the index structure; Set cosine similarity as the similarity metric for comparing text vectors.

8. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 7 is characterized in that: The method utilizes the vector index to retrieve relevant medical literature nodes based on the key information extracted by the question analysis module to obtain corresponding multi-source medical knowledge information, including: Convert the key information extracted by the question parsing module into a query vector; According to the query vector, use the vector index to find the text vector similar to the query vector and calculate the corresponding cosine similarity; The found text vectors are sorted in descending order according to cosine similarity, and the relevant medical documents are returned as multi-source medical knowledge information.

9. The intelligent question-answering system based on multi-source medical knowledge retrieval enhancement according to claim 8, characterized in that: The step of integrating multi-source medical knowledge information to generate multi-source medical knowledge fragments includes: Based on the cosine similarity of the text vectors corresponding to the multi-source medical knowledge information, the top three text vectors ranked in descending order of cosine similarity are extracted to generate multi-source medical knowledge fragments.

Images