Knowledge graph construction method and system

Abstract

The invention discloses a knowledge graph construction method and system, and relates to the field of knowledge graphs, and the method comprises the steps: obtaining medical data, and carrying out the preprocessing of the obtained medical data, and obtaining standard medical data; entity recognition is conducted on the obtained standard medical data, a plurality of entity sets are obtained, and the entity sets comprise a disease name set, a medicine name set, a disease symptom set, an operation name set and a precautions set; constructing a medical knowledge graph according to the obtained entity sets, and perfecting the medical knowledge graph according to the relationship between the entity sets; according to the method, the relevance among the medical entities is fully reflected, the relationship among the medical entities is mined, the medical entity relationship prediction can be more comprehensively and accurately carried out, and then the medical knowledge graph can be more completely constructed.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graphs, and specifically to a knowledge graph construction method and system. Background Art

[0002] Modern medicine is facing the challenge of massive amounts of data and complex information. Medical knowledge is constantly updated, and new drugs, new therapies, and new diseases are discovered very quickly. Knowledge graphs need to be constantly updated and maintained. Knowledge graphs can integrate medical information scattered from different sources, provide a comprehensive knowledge view, discover implicit relationships and new knowledge through reasoning engines, support clinical decision-making and medical research, and standardize knowledge representation methods to facilitate knowledge sharing and reuse, and promote academic exchanges and cooperation.

[0003] In the existing knowledge graph construction process, the relationship between entities mostly needs to be constructed manually or relies on fixed logic, which is a huge workload and difficult to be universally applicable. How to obtain an entity set, build a medical knowledge graph based on the obtained entity set, and improve the medical knowledge graph based on the relationship between the entity sets is a problem we need to solve. To this end, a knowledge graph construction method and system are provided. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a knowledge graph construction method and system.

[0005] The first objective of the present invention is to develop a knowledge graph construction method.

[0006] The second object of the present invention is to develop a knowledge graph construction system.

[0007] To achieve the above first invention objective, the present invention is implemented through the following technical solution: a knowledge graph construction method, comprising the following steps:

[0008] Step S1: acquiring medical data, and preprocessing the acquired medical data to obtain standard medical data;

[0009] Step S2: performing entity recognition on the obtained standard medical data to obtain a number of entity sets, wherein the entity sets include a disease name set, a drug name set, a disease symptom set, a surgery name set, and a precaution set;

[0010] Step S3: Construct a medical knowledge graph based on the obtained entity set, and improve the medical knowledge graph based on the relationship between the entity sets.

[0011] Furthermore, the process of obtaining medical data and preprocessing the obtained medical data includes:

[0012] Obtaining medical data from electronic health records, medical research papers, and medical databases, wherein the medical data includes medical records of several patients;

[0013] Compare the acquired medical data in pairs. If the same medical data exists, remove the duplicate medical data. If the same medical data does not exist, scan the acquired medical data to determine whether there is medical data with missing values. If there is no medical data with missing values, the preprocessing is completed, and the preprocessed medical data is recorded as standard medical data.

[0014] If there is medical data containing missing values, the medical data containing missing values ​​are subjected to secondary preprocessing.

[0015] Furthermore, the process of secondary preprocessing of medical data with missing values ​​includes:

[0016] The medical data containing missing values ​​is deleted, and the medical data containing missing values ​​is acquired for a second time. The medical data acquired for the second time is scanned to determine whether the medical data acquired for the second time contains missing values. If the medical data acquired for the second time contains missing values, the medical data acquired for the second time is deleted. If the medical data acquired for the second time does not contain missing values, the secondary preprocessing is completed.

[0017] Furthermore, the process of entity recognition on the obtained standard medical data includes:

[0018] Performing entity recognition on standard medical data through natural language processing tools to obtain a number of medical entities, wherein the medical entities include disease names, drug names, disease symptoms, surgery names, and precautions, wherein the precautions include precautions for drugs and precautions for surgery;

[0019] Establishing a disease name set, a drug name set, a disease symptom set, a surgery name set, and a precaution set, wherein the precaution set is composed of a drug precaution subset and a surgery precaution subset;

[0020] Map the obtained disease names to the established disease name set, map the obtained drug names to the established drug name set, map the obtained disease symptoms to the established disease symptom set, map the obtained surgery names to the established surgery name set, map the obtained drug precautions to the established drug precautions subset, and map the obtained surgery precautions to the established surgery precautions subset.

[0021] Furthermore, the process of constructing a medical knowledge graph based on the obtained entity set includes:

[0022] The medical knowledge graph is composed of disease name nodes, disease symptom nodes, drug name nodes, surgery name nodes, drug precaution nodes, and surgery precaution nodes;

[0023] Connect the disease name node with the disease symptom node, connect the disease name node with the drug name node, connect the disease name node with the surgery name node, connect the drug name node with the drug precaution node, and connect the surgery name node with the surgery precaution node;

[0024] Map the disease name set to the disease name node, map the disease symptom set to the disease symptom node, map the drug name set to the drug name node, map the surgery name set to the surgery name node, map the drug precautions subset to the drug precautions node, and map the surgery precautions subset to the surgery precautions node.

[0025] Furthermore, the process of improving the medical knowledge graph based on the relationship between entity sets includes:

[0026] Obtaining a connection closeness coefficient between a disease name and a disease symptom, setting a connection closeness coefficient threshold range between the disease name and the disease symptom according to actual conditions, and obtaining the connection closeness between the disease name and the disease symptom according to the obtained connection closeness coefficient between the disease name and the disease symptom and the set connection closeness coefficient threshold range between the disease name and the disease symptom, wherein the connection closeness between the disease name and the disease symptom includes a primary connection closeness, a secondary connection closeness, and a tertiary connection closeness;

[0027] Virtually connect the disease names and disease symptoms of the third-level closeness, and highlight the virtual connection lines between the disease names and disease symptoms of the third-level closeness in green;

[0028] Virtually connect the disease name and disease symptoms of the second-level closeness, and highlight the virtual connection line between the disease name and disease symptoms of the second-level closeness in orange;

[0029] Virtually connect the disease name and disease symptoms of the first-level closeness, and highlight the virtual connection line between the disease name and disease symptoms of the first-level closeness in red;

[0030] Obtain the connection closeness coefficient between the disease name and the drug name, set the connection closeness coefficient threshold range between the disease name and the drug name according to the actual situation, and obtain the connection closeness between the disease name and the drug name according to the obtained connection closeness coefficient between the disease name and the drug name and the set connection closeness coefficient threshold range between the disease name and the drug name, wherein the connection closeness between the disease name and the drug name includes primary connection closeness, secondary connection closeness and tertiary connection closeness;

[0031] Virtually connect the disease names and drug names with the third-level closeness of connection, and highlight the virtual connection lines between the disease names and drug names with the third-level closeness of connection in green;

[0032] Virtually connect the disease name and the drug name of the second-level closeness, and highlight the virtual connection line between the disease name and the drug name of the second-level closeness in orange;

[0033] Virtually connect the disease name and the drug name with the first-level closeness, and highlight the virtual connection line between the disease name and the drug name with the first-level closeness in red;

[0034] By analogy, the connection density between the disease name and the surgery name is obtained, and the connection density between the disease name and the surgery name includes the first-level connection density, the second-level connection density and the third-level connection density, and the disease name and surgery name with the corresponding connection density are virtually connected.

[0035] Furthermore, the process of obtaining the correlation coefficient between the disease name and the disease symptoms includes:

[0036] According to standard medical data, the number of medical records in which the disease name and disease symptoms appear at the same time is obtained, and the total number of medical records is obtained;

[0037] According to the number of medical records in which the disease name and the disease symptoms appear at the same time and the total number of medical records, the connection closeness coefficient between the disease name and the disease symptoms is obtained.

[0038] Based on the same inventive principle, in order to achieve the above second invention objective, the present invention provides a knowledge graph construction system, which adopts a knowledge graph construction method as described in the first invention, and includes the following modules:

[0039] Data acquisition module: used to collect medical data and pre-process the collected medical data to obtain standard medical data;

[0040] A data processing module, used for performing entity recognition on the obtained standard medical data to obtain a plurality of entity sets, wherein the entity sets include a disease name set, a drug name set, a disease symptom set, an operation name set, and a precaution set;

[0041] The knowledge graph construction module is used to construct a medical knowledge graph based on the obtained entity set and to improve the medical knowledge graph based on the relationship between the entity sets.

[0042] Compared with the prior art, the beneficial effects of the present invention are: obtaining medical data, preprocessing the obtained medical data, obtaining standard medical data, performing entity recognition on the obtained standard medical data, obtaining several entity sets, wherein the entity sets include a disease name set, a drug name set, a disease symptom set, a surgery name set, and a precaution set, constructing a medical knowledge graph based on the obtained entity sets, and improving the medical knowledge graph based on the relationships between the entity sets, fully reflecting the correlation between the various medical entities, mining the relationships between the medical entities, being able to more comprehensively and accurately predict the relationships between medical entities, and thus being able to more completely construct a medical knowledge graph. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0044] like Figure 1 As shown, a knowledge graph construction method includes the following steps:

[0045] Step S1: acquiring medical data, and preprocessing the acquired medical data to obtain standard medical data;

[0046] Step S2: performing entity recognition on the obtained standard medical data to obtain a number of entity sets, wherein the entity sets include a disease name set, a drug name set, a disease symptom set, a surgery name set, and a precaution set;

[0047] Step S3: construct a medical knowledge graph based on the obtained entity set, and improve the medical knowledge graph based on the relationship between the entity sets;

[0048] It should be further explained that, in the specific implementation process, the process of obtaining medical data and preprocessing the obtained medical data includes:

[0049] Acquire medical data from electronic health records, medical research papers, medical databases, medical textbooks, pharmaceutical companies, and medical laboratories, wherein the medical data includes medical records of several patients, and pre-process the acquired medical data;

[0050] The acquired medical data are compared with each other. If the same medical data exists, the duplicate medical data are removed. If the same medical data does not exist, the acquired medical data are scanned to determine whether there is medical data containing missing values. If there is no medical data containing missing values, the preprocessing is completed and the preprocessed medical data is recorded as standard medical data. If there is medical data containing missing values, the medical data containing missing values ​​are deleted, and the medical data containing missing values ​​are acquired for the second time. The medical data acquired for the second time is scanned to determine whether the medical data acquired for the second time contains missing values. If the medical data acquired for the second time contains missing values, the medical data acquired for the second time is deleted. If the medical data acquired for the second time does not contain missing values, the preprocessing is completed and the preprocessed medical data is recorded as standard medical data.

[0051] It should be further explained that, in the specific implementation process, the process of entity recognition of the obtained standard medical data includes:

[0052] Performing entity recognition on standard medical data through natural language processing tools to obtain a number of medical entities, wherein the medical entities include disease names, drug names, disease symptoms, surgery names, and precautions, wherein the precautions include precautions for drugs and precautions for surgery;

[0053] Establishing a disease name set, a drug name set, a disease symptom set, a surgery name set, and a precaution set, wherein the precaution set is composed of a drug precaution subset and a surgery precaution subset;

[0054] Map the obtained disease names to the established disease name set, map the obtained drug names to the established drug name set, map the obtained disease symptoms to the established disease symptom set, map the obtained surgery names to the established surgery name set, map the obtained drug precautions to the established drug precautions subset, and map the obtained surgery precautions to the established surgery precautions subset.

[0055] It should be further explained that, in the specific implementation process, the process of constructing a medical knowledge graph based on the obtained entity set includes:

[0056] The medical knowledge graph is composed of disease name nodes, disease symptom nodes, drug name nodes, surgery name nodes, drug precaution nodes, and surgery precaution nodes;

[0057] Connect the disease name node with the disease symptom node, connect the disease name node with the drug name node, connect the disease name node with the surgery name node, connect the drug name node with the drug precaution node, and connect the surgery name node with the surgery precaution node;

[0058] Map the disease name set to the disease name node, map the disease symptom set to the disease symptom node, map the drug name set to the drug name node, map the surgery name set to the surgery name node, map the drug precautions subset to the drug precautions node, and map the surgery precautions subset to the surgery precautions node.

[0059] It should be further explained that, in the specific implementation process, the process of improving the medical knowledge graph according to the relationship between entity sets includes:

[0060] Obtain the correlation coefficient between the disease name and the disease symptoms;

[0061] It should be further explained that, in the specific implementation process, the process of obtaining the correlation coefficient between the disease name and the disease symptoms includes:

[0062] The disease names are numbered as i, i = 1, 2, 3, ..., n, where n is a positive integer. The disease symptoms are numbered as j, j = 1, 2, 3, ..., m, where m is a positive integer. The number of medical records in which the disease name and disease symptoms appear at the same time is obtained according to the standard medical data. The number of medical records in which the disease name and disease symptoms appear at the same time is recorded as C. ij , obtain the total number of medical records, and record the total number of medical records obtained as S;

[0063] The connection closeness coefficient between the disease name and the disease symptom is obtained according to the number of medical records in which the disease name and the disease symptom appear at the same time, and the connection closeness coefficient between the disease name and the disease symptom is recorded as L ij ;

[0064] in, TJ_L is the adjustment coefficient, and TJ_L is a positive number;

[0065] According to the actual situation, the threshold range of the connection closeness coefficient between the disease name and the disease symptom is set, and the set threshold range of the connection closeness coefficient between the disease name and the disease symptom is recorded as (L0, L1);

[0066] According to the obtained connection closeness coefficient between the disease name and the disease symptom and the set connection closeness coefficient threshold range between the disease name and the disease symptom, the connection closeness between the disease name and the disease symptom is obtained, wherein the connection closeness between the disease name and the disease symptom includes a first-level connection closeness, a second-level connection closeness and a third-level connection closeness;

[0067] It should be further explained that, in the specific implementation process, the third-level contact closeness is higher than the second-level contact closeness, which is higher than the first-level contact closeness;

[0068] When 0<L ij When ≤L0, the closeness of the connection between the disease name and the disease symptoms is the first-level closeness;

[0069] When L0<L ij When <L1, the connection density between the disease name and the disease symptoms is the second-level connection density;

[0070] When L ij When ≥L1, the closeness of the connection between the disease name and the disease symptoms is the third-level closeness of the connection;

[0071] Virtually connect the disease names and disease symptoms of the third-level closeness, and highlight the virtual connection lines between the disease names and disease symptoms of the third-level closeness in green;

[0072] Virtually connect the disease name and disease symptoms of the second-level closeness, and highlight the virtual connection line between the disease name and disease symptoms of the second-level closeness in orange;

[0073] Virtually connect the disease name and disease symptoms of the first-level closeness, and highlight the virtual connection line between the disease name and disease symptoms of the first-level closeness in red;

[0074] Get the connection closeness coefficient between the disease name and the drug name;

[0075] It should be further explained that, in the specific implementation process, the process of obtaining the connection closeness coefficient between the disease name and the drug name includes:

[0076] The drug names are labeled as k, k = 1, 2, 3, ..., z, where z is a positive integer. The number of medical records in which the disease name and the drug name appear at the same time is obtained according to the standard medical data, and the number of medical records in which the disease name and the drug name appear at the same time is recorded as C ik药 ;

[0077] The connection closeness coefficient between the disease name and the drug name is obtained according to the number of medical records in which the disease name and the drug name appear at the same time, and the connection closeness coefficient between the disease name and the drug name is recorded as L ik药 ;

[0078] in, TJ_P is the adjustment coefficient, and TJ_P is a positive number;

[0079] According to the actual situation, the threshold range of the connection closeness coefficient between the disease name and the drug name is set, and the set threshold range of the connection closeness coefficient between the disease name and the drug name is recorded as (L2, L3);

[0080] According to the obtained connection closeness coefficient between the disease name and the drug name and the set connection closeness coefficient threshold range between the disease name and the drug name, the connection closeness between the disease name and the drug name is obtained, and the connection closeness between the disease name and the drug name includes a primary connection closeness, a secondary connection closeness and a tertiary connection closeness;

[0081] When 0<L ik药 When ≤L2, the closeness of the connection between the disease name and the drug name is the first-level closeness;

[0082] When L2<L ik药 When <L3, the connection density between the disease name and the drug name is the second-level connection density;

[0083] When L ik药 When ≥L3, the connection density between the disease name and the drug name is the third-level connection density;

[0084] Virtually connect the disease names and drug names with the third-level closeness of connection, and highlight the virtual connection lines between the disease names and drug names with the third-level closeness of connection in green;

[0085] Virtually connect the disease name and the drug name of the second-level closeness, and highlight the virtual connection line between the disease name and the drug name of the second-level closeness in orange;

[0086] Virtually connect the disease name and the drug name with the first-level closeness, and highlight the virtual connection line between the disease name and the drug name with the first-level closeness in red;

[0087] By analogy, the connection closeness coefficient between the disease name and the surgery name is obtained, and the connection closeness between the disease name and the surgery name is obtained according to the obtained connection closeness coefficient between the disease name and the surgery name, wherein the connection closeness between the disease name and the surgery name includes the first-level connection closeness, the second-level connection closeness and the third-level connection closeness;

[0088] Virtually connect the disease names and surgical operation names with the third-level closeness of connection, and highlight the virtual connection lines between the disease names and surgical operation names with the third-level closeness of connection in green;

[0089] Virtually connect the disease name and the surgery name of the second-level closeness of connection, and highlight the virtual connection line between the disease name and the surgery name of the second-level closeness of connection in orange;

[0090] Virtually connect the disease name and the surgery name with the first-level closeness, and highlight the virtual connection line between the disease name and the surgery name with the first-level closeness in red;

[0091] Obtaining the precautions of the drug corresponding to the drug name according to the standard medical data, virtually connecting the drug name with the corresponding precautions of the drug, and highlighting the virtual connection line between the drug name and the corresponding precautions of the drug in green;

[0092] The surgical precautions corresponding to the surgical name are obtained according to standard medical data, the surgical name and the corresponding surgical precautions are virtually connected, and the virtual connection line between the surgical name and the corresponding surgical precautions is highlighted in green.

[0093] The present invention also provides a knowledge graph construction system, which includes:

[0094] Data acquisition module: used to collect medical data and pre-process the collected medical data to obtain standard medical data;

[0095] A data processing module, used for performing entity recognition on the obtained standard medical data to obtain a plurality of entity sets, wherein the entity sets include a disease name set, a drug name set, a disease symptom set, an operation name set, and a precaution set;

[0096] The knowledge graph construction module is used to construct a medical knowledge graph based on the obtained entity set and to improve the medical knowledge graph based on the relationship between the entity sets.

[0097] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A knowledge graph construction method, characterized in that: The following steps are involved: Step S1: acquiring medical data, and preprocessing the acquired medical data to obtain standard medical data; Step S2: performing entity recognition on the obtained standard medical data to obtain a number of entity sets, wherein the entity sets include a disease name set, a drug name set, a disease symptom set, a surgery name set, and a precaution set; Step S3: Construct a medical knowledge graph based on the obtained entity set, and improve the medical knowledge graph based on the relationship between the entity sets.

2. A knowledge graph construction method according to claim 1, characterized in that: The process of acquiring medical data and preprocessing the acquired medical data includes: Obtaining medical data from electronic health records, medical research papers, and medical databases, wherein the medical data includes medical records of several patients; Compare the acquired medical data in pairs. If the same medical data exists, remove the duplicate medical data. If the same medical data does not exist, scan the acquired medical data to determine whether there is medical data with missing values. If there is no medical data with missing values, the preprocessing is completed, and the preprocessed medical data is recorded as standard medical data. If there is medical data containing missing values, the medical data containing missing values ​​are subjected to secondary preprocessing.

3. A knowledge graph construction method according to claim 2, characterized in that: The process of secondary preprocessing of medical data with missing values ​​includes: The medical data containing missing values ​​is deleted, and the medical data containing missing values ​​is acquired for a second time. The medical data acquired for the second time is scanned to determine whether the medical data acquired for the second time contains missing values. If the medical data acquired for the second time contains missing values, the medical data acquired for the second time is deleted. If the medical data acquired for the second time does not contain missing values, the secondary preprocessing is completed.

4. A knowledge graph construction method according to claim 3, characterized in that: The process of entity recognition on the obtained standard medical data includes: Performing entity recognition on standard medical data through natural language processing tools to obtain a number of medical entities, wherein the medical entities include disease names, drug names, disease symptoms, surgery names, and precautions, wherein the precautions include precautions for drugs and precautions for surgery; Establishing a disease name set, a drug name set, a disease symptom set, a surgery name set, and a precaution set, wherein the precaution set is composed of a drug precaution subset and a surgery precaution subset; Map the obtained disease names to the established disease name set, map the obtained drug names to the established drug name set, map the obtained disease symptoms to the established disease symptom set, map the obtained surgery names to the established surgery name set, map the obtained drug precautions to the established drug precautions subset, and map the obtained surgery precautions to the established surgery precautions subset.

5. A knowledge graph construction method according to claim 4, characterized in that: The process of constructing a medical knowledge graph based on the obtained entity set includes: The medical knowledge graph is composed of disease name nodes, disease symptom nodes, drug name nodes, surgery name nodes, drug precaution nodes, and surgery precaution nodes; Connect the disease name node with the disease symptom node, connect the disease name node with the drug name node, connect the disease name node with the surgery name node, connect the drug name node with the drug precaution node, and connect the surgery name node with the surgery precaution node; Map the disease name set to the disease name node, map the disease symptom set to the disease symptom node, map the drug name set to the drug name node, map the surgery name set to the surgery name node, map the drug precautions subset to the drug precautions node, and map the surgery precautions subset to the surgery precautions node.

6. A knowledge graph construction method according to claim 5, characterized in that: The process of improving the medical knowledge graph based on the relationship between entity sets includes: Obtaining a connection closeness coefficient between a disease name and a disease symptom, setting a connection closeness coefficient threshold range between the disease name and the disease symptom according to actual conditions, and obtaining the connection closeness between the disease name and the disease symptom according to the obtained connection closeness coefficient between the disease name and the disease symptom and the set connection closeness coefficient threshold range between the disease name and the disease symptom, wherein the connection closeness between the disease name and the disease symptom includes a primary connection closeness, a secondary connection closeness, and a tertiary connection closeness; Virtually connect the disease names and disease symptoms of the third-level closeness, and highlight the virtual connection lines between the disease names and disease symptoms of the third-level closeness in green; Virtually connect the disease name and disease symptoms of the second-level closeness, and highlight the virtual connection line between the disease name and disease symptoms of the second-level closeness in orange; Virtually connect the disease name and disease symptoms of the first-level closeness, and highlight the virtual connection line between the disease name and disease symptoms of the first-level closeness in red; Obtain the connection closeness coefficient between the disease name and the drug name, set the connection closeness coefficient threshold range between the disease name and the drug name according to the actual situation, and obtain the connection closeness between the disease name and the drug name according to the obtained connection closeness coefficient between the disease name and the drug name and the set connection closeness coefficient threshold range between the disease name and the drug name, wherein the connection closeness between the disease name and the drug name includes primary connection closeness, secondary connection closeness and tertiary connection closeness; Virtually connect the disease names and drug names with the third-level closeness of connection, and highlight the virtual connection lines between the disease names and drug names with the third-level closeness of connection in green; Virtually connect the disease name and the drug name of the second-level closeness, and highlight the virtual connection line between the disease name and the drug name of the second-level closeness in orange; Virtually connect the disease name and the drug name with the first-level closeness, and highlight the virtual connection line between the disease name and the drug name with the first-level closeness in red; By analogy, the connection density between the disease name and the surgery name is obtained, and the connection density between the disease name and the surgery name includes the first-level connection density, the second-level connection density and the third-level connection density, and the disease name and surgery name with the corresponding connection density are virtually connected.

7. A knowledge graph construction method according to claim 6, characterized in that: The process of obtaining the correlation coefficient between the disease name and the disease symptoms includes: According to standard medical data, the number of medical records in which the disease name and disease symptoms appear at the same time is obtained, and the total number of medical records is obtained; According to the number of medical records in which the disease name and the disease symptoms appear at the same time and the total number of medical records, the connection closeness coefficient between the disease name and the disease symptoms is obtained.

8. A knowledge graph construction system, characterized in that: A knowledge graph construction method as described in any one of claims 1 to 7 is adopted, comprising the following modules: Data acquisition module: used to collect medical data and pre-process the collected medical data to obtain standard medical data; A data processing module, used for performing entity recognition on the obtained standard medical data to obtain a plurality of entity sets, wherein the entity sets include a disease name set, a drug name set, a disease symptom set, an operation name set, and a precaution set; The knowledge graph construction module is used to construct a medical knowledge graph based on the obtained entity set and to improve the medical knowledge graph based on the relationship between the entity sets.

Images