Semantic search method and system for medical knowledge base, medium and equipment

Abstract

The invention provides a medical knowledge base semantic search method and system, a medium and equipment. The method comprises the steps of obtaining a natural language query text input by a user; identifying the natural language query text to obtain key information; mapping the key information to corresponding entity nodes and relationships thereof in a knowledge graph to obtain a mapping result; the knowledge graph is constructed based on medical knowledge; and sorting the mapping results, and obtaining a search result according to the sorting. According to the method, the key information in the text is obtained by recognizing the natural language query text input by the user, so that compared with an existing search engine, the natural language query text of the user is analyzed more thoroughly, and the search intention of the user can be understood more deeply. Furthermore, according to the application, the mapping results are sorted, and the mapping results with low association degree with the search intention of the user are excluded, so that the accuracy of the search results is improved.

Description

Technical Field

[0001] The present invention relates to the field of semantic search technology, and in particular to a medical knowledge base semantic search method, system, medium and device. Background Art

[0002] In the current medical industry, the construction and application of knowledge bases are becoming increasingly important, because the rapid development of medical knowledge and the increasing amount of information require medical institutions to be able to efficiently and accurately manage and use this knowledge. The core purpose of the medical knowledge base is to centrally store and organize medical information, including disease diagnosis, treatment methods, drug information, the latest medical research results, etc., so that doctors and medical professionals can quickly obtain the required information and improve the efficiency and quality of diagnosis and treatment.

[0003] However, faced with the increasing amount of medical resource information on the Internet, the existing search methods do not take semantic information into consideration and cannot understand the user's query intent, resulting in low user search efficiency and search accuracy, and cannot provide users with accurate and personalized search services.

[0004] Therefore, a medical knowledge base semantic search method is needed to solve the defects of the existing technology, such as difficulty in understanding user query intent and low relevance of search results. Summary of the invention

[0005] In view of the defects in the prior art, the object of the present invention is to provide a medical knowledge base semantic search method, system, medium and equipment.

[0006] A semantic search method for a medical knowledge base provided according to the present invention includes: obtaining a natural language query text input by a user; identifying the natural language query text to obtain key information; mapping the key information to corresponding entity nodes and their relationships in a knowledge graph to obtain a mapping result; the knowledge graph is constructed based on medical knowledge; and sorting the mapping results to obtain search results based on the sorting.

[0007] It is not difficult to understand that the present application obtains key information in the text by identifying the natural language query text input by the user. Therefore, compared with the existing search engines, the present application analyzes the user's natural language query text more thoroughly and can understand the user's search intent more deeply. Furthermore, the present application sorts the mapping results and excludes the mapping results with low relevance to the user's search intent, thereby improving the accuracy of the search results.

[0008] Furthermore, the intention of identifying the natural language query text includes: the identification of the natural language query text to obtain key information includes: performing named entity recognition on the natural language query text to obtain named entities; performing semantic recognition on the named entities to determine the meanings of the named entities; performing syntactic analysis on the named entities to obtain semantic relationships between the named entities; performing keyword weight analysis on the named entities to determine the weights of the named entities; screening the named entities according to the weights of the named entities to determine the screened named entities; and obtaining key information based on the screened named entities, the meanings of the named entities, and the semantic relationships between the named entities.

[0009] It is not difficult to understand that this application performs named entity recognition on the natural language query text input by the user to find the key named entities in the text, which is conducive to enhancing the extraction of key points in the natural language query text. This application identifies the meaning of the named entities themselves and the semantic relationship between named entities. When faced with entities with multiple meanings, their true meaning can be accurately understood to avoid ambiguity. This application analyzes the weights of named entities to help understand the user's intentions.

[0010] Furthermore, the method for semantically identifying named entities and determining the meaning of named entities includes: using a trained word embedding model to determine the word embedding vector of the named entity; using a trained context embedding model to generate a context embedding vector for each named entity based on the contextual dynamic information of the sentence where the named entity is located; inputting the word embedding vector of the named entity and the context embedding vector of the named entity into a trained convolutional neural network to obtain a representation vector that integrates external knowledge of the text, and obtaining the meaning of each of the named entities based on the representation vector.

[0011] Furthermore, the method of performing syntactic analysis on named entities to obtain semantic relationships between named entities includes: performing dependency analysis on named entities based on transfer dependency analysis to find the core words of the natural language query text where the named entities are located; recursively searching for all words that are parallel to the core words to form a predicate list; for each predicate, finding its subject and object based on the dependency relationship to form an information tuple; for an information tuple without a subject, finding a parent tuple that is parallel to its predicate and has a subject, and if the parent tuple exists, setting the subject of the information tuple to the subject of the parent tuple, and if not found, not processing; returning the processing result to obtain the semantic relationship between the named entities.

[0012] Furthermore, the keyword weight analysis of the named entity to determine the weight of the named entity includes: obtaining the word occurrence rate of the named entity in the natural language query text; obtaining the inverse document frequency of the named entity in the entire document collection; and determining the weight of the named entity based on the occurrence frequency of the named entity in the natural language query text and the inverse document frequency of the named entity in the entire document collection.

[0013] Furthermore, the method of mapping the key information to the corresponding entity nodes and their relationships in the knowledge graph to obtain the mapping results includes:

[0014] Retrieving nodes in the knowledge graph that match the named entities in the key information;

[0015] Calculate the similarity between the named entities in the key information and the entity nodes in the knowledge graph;

[0016] According to the similarity calculation result, the entity nodes whose similarity is greater than a preset value are determined as candidate entity nodes;

[0017] According to the selected candidate entity nodes, the correspondence between the candidate entity nodes is found in the knowledge graph, and the correspondence between the candidate entity nodes with the same semantic relationship as the key information and the candidate entity nodes are selected as the final mapping results.

[0018] It is not difficult to understand that the present application sorts the similarities between named entities and entity nodes, and outputs entity nodes with high similarity and corresponding paths, thereby solving the defect of low correlation of search results existing in existing search engines.

[0019] Furthermore, sorting the mapping results includes sorting the mapping results according to the similarity between the named entities and the entity nodes and the user preference.

[0020] A medical knowledge base semantic search system provided by the present invention comprises:

[0021] An acquisition module is used to acquire the natural language query text input by the user;

[0022] A natural language processing module recognizes the natural language query text to obtain key information;

[0023] A query mapping module maps key information to corresponding entity nodes and their relationships in a knowledge graph to obtain a mapping result; the knowledge graph is constructed based on medical knowledge;

[0024] The ranking algorithm module sorts the mapping results and obtains the search results based on the sorting.

[0025] According to a computer-readable storage medium storing a computer program provided by the present invention, the computer program implements the steps of the medical knowledge base semantic search method when executed by a processor.

[0026] An electronic device provided according to the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the medical knowledge base semantic search method are implemented.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] It is not difficult to understand that the present application obtains key information in the text by identifying the natural language query text input by the user. Therefore, compared with the existing search engines, the present application analyzes the user's natural language query text more thoroughly and can understand the user's search intent more deeply. Furthermore, the present application sorts the mapping results and excludes the mapping results with low relevance to the user's search intent, thereby improving the accuracy of the search results. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0030] Figure 1 This is a flow chart of the semantic search method for medical knowledge base of the present invention;

[0031] Figure 2 It is a flow chart of the syntactic analysis method in the present invention;

[0032] Figure 3 It is a flowchart of the similarity matching between entity nodes and knowledge graphs in the present invention;

[0033] Figure 4 This is the architecture diagram of the medical knowledge base semantic search system of the present invention. DETAILED DESCRIPTION

[0034] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0035] This application provides a method for semantic search of medical knowledge base, referring to Figure 1As shown, it includes: obtaining natural language query text input by a user; identifying the natural language query text to obtain key information; mapping the key information to corresponding entity nodes and their relationships in a knowledge graph to obtain mapping results; the knowledge graph is constructed based on medical knowledge; sorting the mapping results, and obtaining search results based on the sorting.

[0036] It is not difficult to understand that the present application obtains key information in the text by identifying the natural language query text input by the user. Therefore, compared with the existing search engines, the present application analyzes the user's natural language query text more thoroughly and can understand the user's search intent more deeply. Furthermore, the present application sorts the mapping results and excludes the mapping results with low relevance to the user's search intent, thereby improving the accuracy of the search results.

[0037] like Figure 1 As shown, the present application provides a medical knowledge base semantic search method, comprising:

[0038] Step 1: Get the natural language query text entered by the user.

[0039] The natural language query text can be a sentence, a piece of content, or a word entered by the user.

[0040] Step 2: Perform intent recognition on the natural language query text to obtain key information.

[0041] Specifically, the method for intent recognition includes:

[0042] Step 201: Perform named entity recognition on the natural language query text to obtain named entities.

[0043] Named entities can be symptoms, diagnosis results, treatments, drugs, dosages, etc. The recognition of named entities in natural language query texts can be completed using conditional random fields (CRF), support vector machines (SVM), recurrent neural networks (RNN) and long short-term memory networks (LSTM) models, and the present invention does not limit this. Assuming that the user inputs: "Zhang San works in the hospital", through named entity recognition, the entities "Zhang San", "working in..." and "hospital" can be obtained.

[0044] Step 202: Perform semantic recognition on the named entity to determine the meaning of the named entity.

[0045] Semantic recognition can be achieved through word embedding (such as Word2Vec or GloVe) and context embedding (such as BERT or ELMo). Semantic recognition obtains the correct meaning of named entities in natural language query text, eliminating the situation where a named entity corresponds to multiple meanings. For example, "Zhang San" in "Zhang San works in the hospital" is a person's name, not a place name.

[0046] Step 203: Perform syntactic analysis on the named entities to obtain semantic relations between the named entities. The semantic relations may be subject-verb-object structures, which are helpful for understanding the organization of sentences and grammatical rules.

[0047] Step 204: Perform keyword weight analysis on the named entities to determine the weights of the named entities in the natural language query text, filter the named entities according to the weights, and obtain key information, which includes a combination of the filtered named entities, the meanings of the filtered named entities, and the semantic relationships between the filtered named entities.

[0048] In a specific implementation, taking "Zhang San works in a hospital" as an example, the specific implementation process of semantic recognition in step 202 is as follows:

[0049] Step 2021: Use the trained word embedding model to map the named entity to a vector of fixed dimension.

[0050] For example, the embedding vector for each word might look like this (simplified representation):

[0051] "Zhang San": [0.1,-0.2,0.3,...];

[0052] "at": [-0.2,0.4,-0.1,...];

[0053] "hospital": [0.3,0.1,0.2,...];

[0054] "Work": [0.2,0.3,-0.4,...].

[0055] Step 222: Using the trained context embedding model, dynamically generate an embedding vector for each named entity based on the context of the sentence; for example:

[0056] "Zhang San": [0.15,-0.18,0.27,...];

[0057] "in": [-0.17,0.38,-0.09,...];

[0058] "hospital": [0.32, 0.09, 0.21, ...];

[0059] “Work”: [0.22,0.32,-0.36,...].

[0060] Step 2023: Use the word embedding vector output by the word embedding model and the context embedding vector output by the context embedding model to perform semantic recognition. Specifically, the word embedding vector and the context embedding vector are input into the trained convolutional neural network to obtain a representation vector that integrates external knowledge of the text, and the meaning of each named entity is obtained according to the representation vector. For example: the word embedding vector output by the word embedding model is used to preliminarily understand the basic meaning of each word. For example, "hospital" may be associated with "medical care" in the word embedding. The context embedding vector output by the context embedding model is used to further refine the meaning of the word in combination with the context information. In this example, "hospital" is not only related to "medical care", but also a place where people go to see a doctor. "Going to work" may represent an activity in the word embedding vector. The context embedding vector indicates a close connection with "hospital", indicating that this is a place to work. Combining the information represented by the word embedding vector and the information represented by the context embedding vector, the meaning of each named entity can be accurately understood.

[0061] In a specific embodiment, referring to Figure 2 As shown, the specific implementation process of the syntactic analysis in step 203 is as follows:

[0062] Step 2031: performing dependency analysis on the named entities based on the transferred dependency analysis to find the core words of the natural language query text;

[0063] Step 2032: recursively search for all words that are parallel to the core word to form a predicate list;

[0064] Step 2033: for each predicate, find its subject and object according to the dependency relationship to form an information tuple;

[0065] Step 2034: For an information tuple without a subject, find a parent tuple that is parallel to its predicate and has a subject. If the parent tuple exists, set the subject of the information tuple as the subject of the parent tuple. If not found, do not process it.

[0066] Step 2035: Return the processing result to obtain the semantic relationship between the named entities. For example, in “Apple helps digestion”, apple is the subject, helps is the predicate, and digest is the object.

[0067] In a specific implementation, the method for analyzing keyword weights in step 204 includes:

[0068] Step 2041: Calculate the term frequency (TF); calculate the frequency of occurrence of each named entity in the query text; in “Apples help digestion”, “apple” appears once and “digestion” appears once.

[0069] Step 2042: Calculate the inverse document frequency (IDF) to determine the rarity of each named entity in the entire corpus. A corpus is a collection of text to be queried and other content-related texts, which are extracted from the Internet or input by the user. For example, "apple" appears in many documents, while "digestion" appears relatively rarely. Assume that in the corpus, "apple" appears in 10% of the documents and "digestion" appears in 2% of the documents; the IDF calculation is as follows:

[0070] IDF(Apple) = log(total number of documents / 10%) = log(1000 / 100) = log(10)≈1;

[0071] IDF(digest)=log(total number of documents / 2%)=log(1000 / 20)=log(50)≈1.7.

[0072] S2043: Multiply the term frequency (TF) and inverse document frequency (IDF) of each named entity to obtain a TF-IDF value.

[0073] For example, TF(apple) is 1 and IDF(apple) is 1; TF(digestion) is 1 and IDF(digestion) is 1.7;

[0074] TF-IDF(Apple)=TF×IDF=1×1=1

[0075] TF-IDF(digested) = TF × IDF = 1 × 1.7 = 1.7

[0076] From the calculation results, we can see that digestion has the highest TF-IDF, which means that in this text, "digestion" is the most unique named entity, and therefore, "digestion" has the highest weight in the entire text.

[0077] Using the above method, the weight of each named entity is calculated separately, and the named entities are prioritized according to the weight of each named entity, and the named entities with the top K priorities are screened out. It is not difficult to understand that the greater the weight of a named entity, the higher the priority of the named entity. For example, K is 30, the weight of each named entity is calculated separately, and the named entities are prioritized according to the weight of each named entity. The named entities with the top 30 priorities can be selected as the screened named entities. Based on the screened named entities, combined with their meanings and the semantic relationships between the named entities, key information is obtained.

[0078] It is not difficult to understand that the key information of the present application is a combination of the screened named entities, the meanings of the screened named entities, and the semantic relationships between the screened named entities.

[0079] Step 3: Map the named entities in the key information to the corresponding entity nodes in the knowledge graph to obtain the mapping results. The knowledge graph is constructed based on medical knowledge; in this application, medical knowledge can come from medical literature, databases, expert knowledge, etc., and the present invention does not limit this.

[0080] Specifically, refer to Figure 3 As shown, the key information is mapped to the corresponding entity nodes and their relationships in the knowledge graph, and the method for obtaining the mapping result includes:

[0081] Step 301: Retrieve nodes in the knowledge graph that match the named entities in the key information.

[0082] In a specific embodiment, the search method includes:

[0083] Step 3011: Eliminate entity nodes that are obviously irrelevant to the searched named entity.

[0084] Since there may be multiple entity nodes with the same name in the knowledge graph, we first need to exclude entity nodes that are obviously irrelevant to the named entities in the key information. For example, for the named entity "apple" in the natural language query text "Apples are rich in dietary fiber and help digestion", there may be multiple entities in the knowledge graph, such as "Apple Inc." (a US technology company) or "apple" (fruit). At this time, we need to determine which entity it refers to based on the meaning of the named entity selected in the key information. It is highly likely that the "apple" here should be a fruit, not a technology company.

[0085] Step 3012: Map the named entities in the key information to the corresponding entity nodes in the knowledge graph through entity linking.

[0086] For example, for the text "Apples are rich in dietary fiber and help digestion", the named entities "apple" and "digestion" should be mapped to the corresponding entity nodes. For example, the entity nodes corresponding to the "apple" entity include "apple peel", "fruit", "apple pie", etc.

[0087] Step 302: Calculate the similarity between the named entities in the key information and the entity nodes in the knowledge graph.

[0088] Since the number of entity nodes matched in the knowledge graph is not unique, it is necessary to calculate the similarity between each entity node and the named entity in the natural language query text. The higher the similarity, the higher the relevance of the matched entity node to the final search result. The similarity between the named entity and the candidate entity in the knowledge graph can be calculated by comparing strings, using context information, or using more complex semantic similarity calculation methods.

[0089] For example, in the present application, the similarity between the entities in the natural language query text and the entity nodes in the knowledge graph is calculated by a similarity algorithm; the similarity algorithm can adopt cosine similarity (Cosine Similarity) and adjusted cosine similarity (Adjusted Cosine Similarity), Pearson correlation coefficient (Pearson Correlation Coefficient), Jaccard similarity coefficient (Jaccard Coefficient), Tanimoto coefficient (generalized Jaccard similarity coefficient), log-likelihood similarity / log-likelihood similarity, mutual information / information gain, relative entropy / KL divergence, term frequency-inverse document frequency (TF-IDF) commonly used in information retrieval, and any one of the similarity algorithms based on word vectors and sentence vectors, and can also be other similarity algorithms, which are not limited in the present invention.

[0090] Step 303: According to the similarity calculation result, the entity nodes with similarity greater than a preset value are determined as candidate entity nodes. For example, several entity nodes with the greatest similarity can be used as candidate entity nodes, that is, entity nodes corresponding to the named entities in the key information.

[0091] Step 304: According to the selected candidate entity nodes, the corresponding relationship between the candidate entity nodes is found in the knowledge graph, and the corresponding relationship between the candidate entity nodes with the same semantic relationship as the key information and the candidate entity nodes are selected as the final mapping result.

[0092] Step 4: Sort the mapping results and obtain the search results according to the sorting.

[0093] The sorting method used in this application is based on the similarity between the named entities and the entity nodes and the user preferences. Other sorting methods are also possible, and this application does not limit this.

[0094] Reference Figure 4 As shown, a medical knowledge base semantic search system is used to execute the above-mentioned electronic health record structured processing method. The system includes:

[0095] An acquisition module is used to acquire the natural language query text input by the user;

[0096] A natural language processing module recognizes the natural language query text to obtain key information;

[0097] A query mapping module maps key information to corresponding entity nodes and their relationships in a knowledge graph to obtain a mapping result; the knowledge graph is constructed based on medical knowledge;

[0098] The ranking algorithm module sorts the mapping results and obtains the search results based on the sorting.

[0099] In other embodiments, the medical knowledge base semantic search system also includes a user interface and interaction module, a knowledge graph construction and maintenance module, a user feedback and learning module, a system integration and testing module, and a deployment and maintenance module.

[0100] The user feedback and learning module is used to collect user feedback and continuously optimize the accuracy and relevance of search results. Through the feedback loop mechanism, user feedback is used to train and adjust the NLP model and ranking algorithm, so as to change the order of mapping results. The system integration and testing module is used to ensure that the modules of the system work together, and to perform system testing and performance evaluation. Through the interface standardization and data synchronization mechanism between modules, integration testing and stress testing are carried out. At the same time, the system integration and test results are fed back to each module to guide system optimization and upgrade. The deployment and maintenance module is responsible for the deployment, monitoring and maintenance of the system to ensure the stable operation of the system. Through the integrated monitoring system, the system performance indicators are tracked in real time to realize the automatic deployment and rollback mechanism. At the same time, the system operation data and maintenance logs are collected for continuous improvement of the system.

[0101] It should be understood that the embodiments of the present invention can be implemented in hardware, software or a combination thereof. Software can be stored in a volatile or non-volatile storage device (regardless of whether the data is deletable or rewritable) like a ROM, in a memory like a RAM, a memory chip, a device or an integrated circuit, or in a storage medium on which data can be optically or magnetically recorded and from which data can be read by a machine (e.g., a computer), such as a compact disc (CD), a digital versatile disc (DVD), a disk, or a tape. Moreover, the embodiments of the present invention can be implemented in a computer or portable terminal having a controller and a memory, and the memory is an example of a non-temporary computer-readable storage medium suitable for storing one or more programs including commands for implementing the embodiments of the present invention. Therefore, the present invention includes a program having a code for implementing the device or method defined by the claims, and a machine-readable non-temporary storage medium storing the program. The program can be transferred electrically by a medium such as a communication signal transmitted via a wired or wireless connection, and the program and its equivalent are included in the present invention.

[0102] The device may receive a program from a program providing device connected via a cable or wirelessly and store the program. The program providing device may include a program containing commands for implementing the embodiments of the present invention, a memory for storing information required by the embodiments of the present invention, a communication module for communicating with the mobile device via a cable or wirelessly, and a controller for transmitting the program to the device automatically or upon request of the device.

[0103] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0104] In the description of the present application, it should be understood that the terms "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0105] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A semantic search method for a medical knowledge base, characterized in that: include: Get the natural language query text entered by the user; Recognize the natural language query text to obtain key information; Mapping key information to corresponding entity nodes and their relationships in a knowledge graph to obtain a mapping result; the knowledge graph is constructed based on medical knowledge; The mapping results are sorted and the search results are obtained according to the sorting.

2. The medical knowledge base semantic search method according to claim 1, characterized in that: The key information obtained by identifying the natural language query text includes: Perform named entity recognition on natural language query text to obtain named entities; Performing semantic recognition on the named entity to determine the meaning of the named entity; Performing syntactic analysis on the named entities to obtain semantic relationships between the named entities; Performing keyword weight analysis on the named entity to determine the weight of the named entity; Screening the named entities according to the weights of the named entities to determine the screened named entities; Key information is obtained based on the screened named entities, the meanings of the named entities, and the semantic relationships between the named entities.

3. The medical knowledge base semantic search method according to claim 2, characterized in that: The method for performing semantic recognition on named entities and determining the meaning of named entities includes: Use the trained word embedding model to determine the word embedding vector of the named entity; Using the trained context embedding model, the context embedding vector of each named entity is generated according to the context dynamic information of the sentence where the named entity is located; The word embedding vector of the named entity and the context embedding vector of the named entity are input into a trained convolutional neural network to obtain a representation vector that integrates external knowledge of the text, and the meaning of each named entity is obtained according to the representation vector.

4. The medical knowledge base semantic search method according to claim 2, characterized in that: The method of performing syntactic analysis on named entities to obtain semantic relations between named entities includes: Perform dependency analysis on the named entities based on the transfer dependency analysis to find the core words of the natural language query text where the named entities are located; Recursively search for all words that are parallel to the core word to form a predicate list; For each predicate, find its subject and object according to the dependency relationship to form an information tuple; For an information tuple without a subject, find a parent tuple that is parallel to its predicate and has a subject. If the parent tuple exists, set the subject of the information tuple as the subject of the parent tuple. If not found, do not process it. The processing result is returned to obtain the semantic relationship between the named entities.

5. The medical knowledge base semantic search method according to claim 2, characterized in that: The performing keyword weight analysis on the named entity to determine the weight of the named entity includes: Obtaining the occurrence rate of the named entity in the natural language query text; Get the inverse document frequency of the named entity in the entire document collection; The weight of the named entity is determined based on the occurrence frequency of the named entity in the natural language query text and the inverse document frequency of the named entity in the entire document collection.

6. The medical knowledge base semantic search method according to claim 1, characterized in that: The method of mapping the key information to the corresponding entity nodes and their relationships in the knowledge graph to obtain the mapping results includes: Retrieving nodes in the knowledge graph that match the named entities in the key information; Calculate the similarity between the named entities in the key information and the entity nodes in the knowledge graph; According to the similarity calculation result, the entity nodes whose similarity is greater than a preset value are determined as candidate entity nodes; According to the selected candidate entity nodes, the correspondence between the candidate entity nodes is found in the knowledge graph, and the correspondence between the candidate entity nodes with the same semantic relationship as the key information and the candidate entity nodes are selected as the final mapping results.

7. The medical knowledge base semantic search method according to claim 1, characterized in that: Sorting the mapping results includes sorting the mapping results based on the similarity between the named entities and the entity nodes and the user preferences.

8. A medical knowledge base semantic search system, characterized in that: include: An acquisition module is used to acquire the natural language query text input by the user; A natural language processing module recognizes the natural language query text to obtain key information; A query mapping module maps key information to corresponding entity nodes and their relationships in a knowledge graph to obtain a mapping result; the knowledge graph is constructed based on medical knowledge; The ranking algorithm module sorts the mapping results and obtains the search results based on the sorting.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the medical knowledge base semantic search method described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is executed by a processor, the steps of the medical knowledge base semantic search method described in any one of claims 1 to 7 are implemented.

Images