Medical question and answer recommendation method, recommendation system and computer-readable storage medium

Abstract

The present invention provides a medical question and answer recommendation method, a recommendation system and a computer-readable storage medium. The medical question and answer recommendation method includes: obtaining medical questions; preprocessing the medical questions to determine multiple keywords; constructing a knowledge graph triple set based on multiple keywords; querying a candidate answer set for the medical question in an existing knowledge graph based on the knowledge graph triple set; extracting the best answer from the candidate answer set and pushing it. Through the technical solution of the present invention, it is realized to combine statistical analysis, knowledge graph construction and feature selection, and automatically recommend medical question and answer scenarios under natural language by means of human-computer interaction. While ensuring the accuracy and professionalism of the recommended answers and meeting the needs of users to obtain health and medical related knowledge, it improves the utilization rate of medical information resources and information query efficiency, saving redundant working time for medical workers.

Description

Technical Field

[0001] The present invention relates to the fields of computer technology and medical informationization, and in particular to a medical question and answer recommendation method, a medical question and answer recommendation system and a computer-readable storage medium. Background Art

[0002] With the development of mobile Internet and medical informationization, mobile medical service business and doctor-patient communication services have a relatively broad development prospect. Faced with the growing number of mobile users and the demand for obtaining health and medical related knowledge, the intelligent medical question-and-answer system can not only allow users to ask questions in natural language, but also assist operators to quickly return concise and accurate answers to users, so as to save the investment of operating personnel and improve user experience.

[0003] Medical data has characteristics such as pattern polymorphism (such as multiple patients with the same disease), incompleteness, timeliness, redundancy and privacy, which pose challenges to medical-related natural language processing technologies. Therefore, in order to build a medical question-and-answer recommendation system suitable for medical scenarios, it is necessary to build scenario-specific natural language processing technologies for medical data, especially free question-and-answer text data in the medical field. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art or related art.

[0005] To this end, one aspect of the present invention is to propose a medical question and answer recommendation method.

[0006] Another aspect of the present invention is to provide a medical question and answer recommendation system.

[0007] Yet another aspect of the present invention provides a computer-readable storage medium.

[0008] In view of this, the present invention proposes a medical question and answer recommendation method, including: obtaining medical questions; preprocessing the medical questions to determine multiple keywords; constructing a knowledge graph triple set based on the multiple keywords; querying a set of candidate answers to the medical questions in the existing knowledge graph based on the knowledge graph triple set; extracting the best answer from the candidate answer set and pushing it.

[0009] The medical question and answer recommendation method provided by the present invention obtains the user's medical questions through various forms such as application programming interface (API) and message push, determines multiple keywords in the medical questions and constructs a knowledge graph triple set based on the keywords, obtains the focus of the user's query sentence and the type of the user's question, and then searches for words and phrases that match the medical question in the existing knowledge graph according to the knowledge graph triple set to form a candidate answer set, extracts the best answer from it, and pushes the best answer to the user. Among them, the purpose of obtaining medical question data is divided into two parts. The first is to first obtain a large amount of medical question data, perform word segmentation entity recognition, etc. to construct an existing knowledge graph and answer matching, etc., which is the training process; the second is to obtain a certain medical question through the mobile terminal, extract keywords and then enter the recommended final answer.

[0010] The medical question and answer recommendation method provided by the present invention can combine natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level and understand the medical questions raised by users. At the same time, it can combine statistical analysis, knowledge graph construction and feature selection, and use human-computer interaction to automatically recommend medical question and answer scenarios under natural language. While ensuring the accuracy and professionalism of the recommended answers and meeting the needs of users for obtaining health and medical related knowledge, it improves the utilization rate of medical information resources and information query efficiency, saving redundant working time for medical workers.

[0011] Specifically, the medical question types include: inquiry type, question words, and expected answer type.

[0012] The medical question and answer recommendation method according to the present invention may also have the following technical features:

[0013] In the above technical solution, preferably, the step of preprocessing the medical question to obtain multiple keywords includes: performing word segmentation and entity recognition on the medical question according to the sentence pattern of the medical question and a preset medical corpus to obtain multiple keywords.

[0014] In this technical solution, according to the sentence pattern of medical questions and the preset medical corpus, the inverse maximum matching algorithm and multi-layer conditional random field are used to perform word segmentation and entity recognition on the user's medical questions to determine the keywords. The medical question and answer recommendation method provided by the present invention can be combined with natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level, thereby improving information query efficiency.

[0015] Specifically, the word segmentation tool is optimized according to the characteristics of the medical field, and the medical corpus data structure is defined based on the foreign medical ontology structure.

[0016] In any of the above technical solutions, preferably, the step of preprocessing the medical problem to obtain multiple keywords also includes: performing statistics on the multiple keywords and weighting each of the multiple keywords based on statistical features.

[0017] In this technical solution, multiple keywords are counted and each of the multiple keywords is weighted based on statistical characteristics to obtain keywords with higher relevance, ensure the accuracy of semantic understanding in specific fields, and improve the utilization rate of medical information resources and information query efficiency.

[0018] Specifically, the statistical features include any one of the following or a combination thereof: word frequency, position, and co-occurrence frequency.

[0019] In any of the above technical solutions, preferably, the step of constructing a knowledge graph triple set based on multiple keywords includes: extracting medical entities and entity relationships from multiple keywords, performing synonym conversion, and constructing a knowledge graph triple set in the form of triple storage.

[0020] In this technical solution, in the field of medical question-answering, the authority of experts is very important. Therefore, methods such as conditional random fields are used to extract medical entities and entity relationships from multiple keywords, and synonym conversion is performed. A knowledge graph triple set is constructed in the form of triple storage. By constructing a knowledge graph triple set, keywords are integrated and a resource description framework (RDF) is defined to facilitate understanding of medical questions raised by users and improve the utilization rate of medical information resources and information query efficiency.

[0021] Specifically, the triple set includes: resource, attribute type and attribute value.

[0022] In any of the above technical solutions, preferably, the step of querying a set of candidate answers to medical questions in an existing knowledge graph based on a knowledge graph triple set includes: determining a knowledge graph subgraph based on the weight of each keyword; matching a knowledge graph path based on the knowledge graph triple set; and searching in the knowledge graph subgraph based on the knowledge graph path to query a set of candidate answers to the medical question.

[0023] In this technical solution, considering the complexity of medical knowledge and the large differences in characteristics of various diseases, the knowledge graph subgraph is determined according to the weight of each keyword, the query scope is narrowed, and the knowledge graph path is matched according to the knowledge graph triple set. The medical entity and entity relationship are combined to constrain the retrieval path. The keyword matching technology is used to search in the knowledge graph subgraph according to the knowledge graph path, and the words and phrases closest to the medical problem are selected to form a set of candidate answers. While ensuring the accuracy and professionalism of the recommended answers, the utilization rate of medical information resources and the efficiency of information query are improved, saving redundant working time for medical workers.

[0024] In any of the above technical solutions, preferably, the step of extracting the best answer from the candidate answer set and pushing it includes: screening the candidate answers in the candidate answer set according to the label of each candidate answer, and / or text features and non-text features; sorting one or more screened candidate answers to determine the best answer.

[0025] In this technical solution, the label of each candidate answer is analyzed, and high-quality answers are further screened based on the label and / or text features and non-text features. The maximum entropy method is used to predict the answer quality, the candidate answers are sorted, and the best answer is selected, ensuring the accuracy and professionalism of the recommended answers and meeting the user's needs for obtaining health and medical related knowledge.

[0026] According to the second aspect of the present invention, a medical question and answer recommendation system is proposed, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following are achieved: obtaining medical questions; preprocessing the medical questions to determine multiple keywords; constructing a knowledge graph triple set based on multiple keywords; querying a candidate answer set for the medical question in an existing knowledge graph based on the knowledge graph triple set; determining the answer quality of each candidate answer in the candidate answer set to extract the best answer and push it. Among them, the purpose of obtaining medical question data is divided into two parts. The first is to first obtain a large amount of medical question data, perform word segmentation and entity recognition, etc. to construct an existing knowledge graph and answer matching, etc., which is the training process; the second is to obtain a certain medical question through a mobile terminal, extract keywords, and then enter the recommended final answer.

[0027] The medical question and answer recommendation system provided by the present invention can combine natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level and understand the medical questions raised by users. At the same time, it combines statistical analysis, knowledge graph construction and feature selection, and uses human-computer interaction to automatically recommend medical question and answer scenarios under natural language. While ensuring the accuracy and professionalism of the recommended answers and meeting the needs of users for obtaining health and medical related knowledge, it improves the utilization rate of medical information resources and information query efficiency, saving redundant working time for medical workers.

[0028] In the above technical solution, preferably, when the processor executes the computer program, it specifically implements the step of preprocessing the medical questions to obtain multiple keywords, including: performing word segmentation and entity recognition on the medical questions according to the sentence pattern of the medical questions and a preset medical corpus to obtain multiple keywords.

[0029] In this technical solution, according to the sentence pattern of medical questions and the preset medical corpus, the inverse maximum matching algorithm and multi-layer conditional random field are used to perform word segmentation and entity recognition on the user's medical questions to determine the keywords. The medical question and answer recommendation method provided by the present invention can be combined with natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level, thereby improving information query efficiency.

[0030] Specifically, the word segmentation tool is optimized according to the characteristics of the medical field, and the medical corpus data structure is defined based on the foreign medical ontology structure.

[0031] In any of the above technical solutions, preferably, when the processor executes the computer program, it further specifically implements: performing statistics on the multiple keywords, and weighting each of the multiple keywords based on the statistical features.

[0032] In this technical solution, multiple keywords are counted and each of the multiple keywords is weighted based on statistical characteristics to obtain keywords with higher relevance, ensure the accuracy of semantic understanding in specific fields, and improve the utilization rate of medical information resources and information query efficiency.

[0033] Specifically, the statistical features include any one of the following or a combination thereof: word frequency, position, and co-occurrence frequency.

[0034] In any of the above technical solutions, preferably, when the processor executes the computer program, it specifically implements the steps of constructing a knowledge graph triple set based on multiple keywords, including: extracting medical entities and entity relationships from multiple keywords, performing synonym conversion, and constructing a knowledge graph triple set in the form of triple storage.

[0035] In this technical solution, in the field of medical question answering, the authority of experts is very important. Therefore, methods such as conditional random fields are used to extract medical entities and entity relationships from multiple keywords, and synonym conversion is performed. A knowledge graph triple set is constructed in the form of triple storage. By constructing a knowledge graph triple set, keywords are integrated and a resource description framework (RDF) is defined to facilitate understanding of medical questions raised by users, thereby improving the utilization rate of medical information resources and information query efficiency.

[0036] Specifically, the triple set includes: resource, attribute type and attribute value.

[0037] In any of the above technical solutions, preferably, when the processor executes the computer program, it specifically implements the step of querying a set of candidate answers to medical questions in an existing knowledge graph based on a knowledge graph triple set, including: determining a knowledge graph subgraph based on the weight of each keyword; matching a knowledge graph path based on the knowledge graph triple set; searching in the knowledge graph subgraph based on the knowledge graph path to query a set of candidate answers to the medical question.

[0038] In this technical solution, considering the complexity of medical knowledge and the large differences in characteristics of various diseases, the knowledge graph subgraph is determined according to the weight of each keyword, the query scope is narrowed, and the knowledge graph path is matched according to the knowledge graph triple set. The medical entity and entity relationship are combined to constrain the retrieval path. The keyword matching technology is used to search in the knowledge graph subgraph according to the knowledge graph path, and the words and phrases closest to the medical problem are selected to form a set of candidate answers. While ensuring the accuracy and professionalism of the recommended answers, the utilization rate of medical information resources and the efficiency of information query are improved, saving redundant working time for medical workers.

[0039] In any of the above technical solutions, preferably, when the processor executes the computer program, it specifically implements the steps of extracting the best answer from the candidate answer set and pushing it, including: screening the candidate answers in the candidate answer set according to the label of each candidate answer, and / or text features and non-text features; sorting one or more screened candidate answers to determine the best answer.

[0040] In this technical solution, the label of each candidate answer is analyzed, and high-quality answers are further screened based on the label and / or text features and non-text features. The maximum entropy method is used to predict the answer quality, the candidate answers are sorted, and the best answer is selected, ensuring the accuracy and professionalism of the recommended answers and meeting the user's needs for obtaining health and medical related knowledge.

[0041] According to a third aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of any of the above-mentioned medical question and answer recommendation methods are implemented. Therefore, the computer-readable storage medium includes all the beneficial effects of the medical question and answer recommendation method of any of the above-mentioned technical solutions.

[0042] Additional aspects and advantages of the present invention will become apparent in the following description or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0044] Figure 1 A schematic diagram showing a flow chart of a medical question and answer recommendation method according to an embodiment of the present invention;

[0045] Figure 2 A schematic diagram showing a flow chart of a medical question and answer recommendation method according to another embodiment of the present invention;

[0046] Figure 3 A schematic diagram showing a flow chart of a medical question and answer recommendation method according to another embodiment of the present invention;

[0047] Figure 4 A schematic diagram showing a process of querying a candidate answer set according to an embodiment of the present invention;

[0048] Figure 5 A schematic diagram showing a flow chart of a medical question and answer recommendation method according to another embodiment of the present invention;

[0049] Figure 6 A schematic diagram showing a flow chart of a medical question and answer recommendation method according to a specific embodiment of the present invention;

[0050] Figure 7 A schematic diagram showing a flow chart of a medical question and answer recommendation method according to another specific embodiment of the present invention;

[0051] Figure 8 A schematic block diagram of a medical question and answer recommendation system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0052] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0054] The embodiment of the first aspect of the present invention provides a medical question and answer recommendation method.

[0055] Figure 1 FIG. 1 is a flow chart showing a medical question and answer recommendation method according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0056] Step 102, obtaining medical problems;

[0057] Step 104, pre-processing the medical question to determine multiple keywords;

[0058] Step 106, constructing a knowledge graph triple set based on multiple keywords;

[0059] Step 108, querying a set of candidate answers to the medical question in the existing knowledge graph according to the knowledge graph triple set;

[0060] Step 110: extract the best answer from the candidate answer set and push it.

[0061] The medical question and answer recommendation method provided in this embodiment obtains the user's medical questions through various forms such as application program interface (API) and message push, determines multiple keywords in the medical questions and constructs a knowledge graph triple set based on the keywords, obtains the focus of the user's query sentence and the type of the user's question, and then searches for words and phrases that match the medical question in the existing knowledge graph according to the knowledge graph triple set to form a candidate answer set, extracts the best answer from it, and pushes the best answer to the user. Among them, the purpose of obtaining medical question data is divided into two parts. The first is to first obtain a large amount of medical question data, perform word segmentation and entity recognition, etc. to construct an existing knowledge graph and answer matching, etc. This is the training process; the second is to obtain a certain medical question through the mobile terminal, extract keywords, and then enter the recommended final answer. The medical question and answer recommendation method provided in this embodiment can combine natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level and understand the medical questions raised by users. At the same time, it combines statistical analysis, knowledge graph construction and feature selection, and uses human-computer interaction to automatically recommend medical question and answer scenarios under natural language. While ensuring the accuracy and professionalism of the recommended answers, it improves the utilization rate of medical information resources and information query efficiency, saving redundant working time for medical workers.

[0062] Figure 2FIG. 2 shows a flow chart of a medical question and answer recommendation method according to another embodiment of the present invention. Figure 2 As shown, the method includes:

[0063] Step 202, obtaining medical problems;

[0064] Step 204, performing word segmentation and entity recognition on the medical question according to the sentence pattern of the medical question and a preset medical corpus to obtain a plurality of keywords;

[0065] Step 206, performing statistics on the multiple keywords, and weighting each of the multiple keywords based on the statistical features;

[0066] Step 208, constructing a knowledge graph triple set based on multiple keywords;

[0067] Step 210, querying a set of candidate answers to the medical question in the existing knowledge graph according to the knowledge graph triple set;

[0068] Step 212: extract the best answer from the candidate answer set and push it.

[0069] In this embodiment, according to the sentence pattern of the medical question and the preset medical corpus, the inverse maximum matching algorithm and the multi-layer conditional random field are used to perform word segmentation and entity recognition on the user's medical question, determine the keywords, and perform statistics on multiple keywords. Based on the statistical characteristics, each of the multiple keywords is weighted to obtain keywords with higher relevance. The medical question and answer recommendation method provided by the present invention can combine natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level and ensure the accuracy of semantic understanding in a specific field, thereby improving the utilization rate of medical information resources and information query efficiency.

[0070] In a specific embodiment, the word segmentation tool is optimized according to the characteristics of the medical field, and the medical corpus data structure is defined based on the foreign medical ontology structure. The statistical features include any one of the following or a combination thereof: word frequency, position, and co-occurrence frequency.

[0071] Figure 3 FIG. 2 shows a flow chart of a medical question and answer recommendation method according to another embodiment of the present invention. Figure 3 As shown, the method includes:

[0072] Step 302, obtaining medical problems;

[0073] Step 304, pre-processing the medical question to determine multiple keywords;

[0074] Step 306, extracting medical entities and entity relationships from multiple keywords, performing synonym conversion, and constructing a knowledge graph triple set in the form of triple storage;

[0075] Step 308, querying a set of candidate answers to the medical question in the existing knowledge graph according to the knowledge graph triple set;

[0076] Step 310: extract the best answer from the candidate answer set and push it.

[0077] In this embodiment, in the field of medical question and answer, the authority of experts is very important. Therefore, methods such as conditional random fields are used to extract medical entities and entity relationships from multiple keywords, and synonym conversion is performed. A knowledge graph triple set is constructed in the form of triple storage. By constructing the knowledge graph triple set, keywords are integrated and a resource description framework (RDF) is defined to facilitate understanding of medical questions raised by users, thereby improving the utilization rate of medical information resources and information query efficiency.

[0078] In a specific embodiment, the triple set includes: resource, attribute type and attribute value.

[0079] Furthermore, the above step 210 may be implemented by another embodiment, such as Figure 4 As shown, specifically including:

[0080] Step 402, determining a knowledge graph subgraph according to the weight of each keyword;

[0081] Step 404, matching the knowledge graph path according to the knowledge graph triple set;

[0082] Step 406, search in the knowledge graph subgraph according to the knowledge graph path to query a set of candidate answers to the medical question.

[0083] In this embodiment, considering the complexity of medical knowledge and the large differences in characteristics of various diseases, the knowledge graph subgraph is determined according to the weight of each keyword, the query scope is narrowed, and the knowledge graph path is matched according to the knowledge graph triple set. The medical entity and entity relationship are combined to constrain the retrieval path. The keyword matching technology is used to search in the knowledge graph subgraph according to the knowledge graph path, and the words and phrases closest to the medical problem are selected to form a set of candidate answers. While ensuring the accuracy and professionalism of the recommended answers, the utilization rate of medical information resources and the efficiency of information query are improved, saving redundant working time for medical workers.

[0084] Figure 5 FIG. 2 is a flow chart showing a medical question and answer recommendation method according to another embodiment of the present invention. Figure 5 As shown, the method includes:

[0085] Step 502, obtaining medical problems;

[0086] Step 504, pre-processing the medical question to determine a plurality of keywords;

[0087] Step 506, constructing a knowledge graph triple set based on multiple keywords;

[0088] Step 508, querying a set of candidate answers to the medical question in the existing knowledge graph according to the knowledge graph triple set;

[0089] Step 510, screening candidate answers in the candidate answer set according to the label of each candidate answer, and / or text features and non-text features;

[0090] Step 512: sort the screened one or more candidate answers to determine the best answer.

[0091] In this embodiment, the label of each candidate answer is analyzed, and higher-quality answers are further screened based on the label and / or text features and non-text features. The maximum entropy method is used to predict the answer quality, the candidate answers are ranked, and the best answer is selected, thereby ensuring the accuracy and professionalism of the recommended answers and meeting the user's needs for obtaining health and medical related knowledge.

[0092] Figure 6 FIG. 1 is a flow chart showing a medical question and answer recommendation method according to a specific embodiment of the present invention. Figure 6 As shown, the method includes:

[0093] Step 602, obtaining user questions through the mobile terminal;

[0094] Establish a user human-computer interaction interface on the mobile terminal to obtain user questions;

[0095] Step 604, performing word segmentation and entity recognition to determine keywords;

[0096] Based on basic sentence patterns and self-built medical corpus, we segmented, recognized entities and parsed the questions to obtain new index segmentation sequences. Then, we used shallow syntactic analysis technology to identify question types based on sentence patterns and question classification.

[0097] Step 606, constructing a knowledge graph based on the medical question-answer pairs and the clinical expert knowledge system;

[0098] Use web crawlers and own data to collect medical question-answer data, use conditional random fields and other methods to extract medical entities and entity relationships, and perform synonym conversion. Define the RDF triple storage format to build a medical question-answer knowledge graph, extract knowledge from it, further influence the weight of questions and answers, and ensure the accuracy of semantic understanding in specific fields;

[0099] Step 608, perform information retrieval to select candidate answers;

[0100] Considering the complexity of medical knowledge and the large differences in characteristics of various diseases, it is necessary to combine clinical experience and disease classification, narrow the knowledge graph according to keywords to form a knowledge graph subgraph, and determine the search scope. Then, the retrieval method is implemented based on keyword fuzzy matching, and the knowledge graph path is matched based on the triple model. For better matching, both the question and the answer can be parsed into a grammar book, and the dependency relationship between words can be obtained from it. Keyword search is performed in the knowledge graph subgraph to generate a set of candidate answers;

[0101] Step 610, extracting the best answer from the candidate answers;

[0102] Find words and phrases that match the question to form candidate answers, combine answer labels and non-text features to further screen high-quality answers, and use the maximum entropy method to predict the answer quality, sort the candidate answers, and select the best answer;

[0103] Step 612, using the mobile terminal to return the best answer.

[0104] The best answer is output on the mobile human-computer interaction interface. To support user personalized needs, multiple candidate answers can be output according to the candidate answer sorting.

[0105] Step 602 and step 612 can be input and output to each other. Based on the iteration of steps 604 to 608, recommended questions are generated and pushed to the mobile terminal, and the question and answer recommendation process is started again after selection is made using the human-computer interaction interface.

[0106] Figure 7 FIG. 2 is a flow chart showing a medical question and answer recommendation method according to another specific embodiment of the present invention. Figure 7 As shown, the method includes:

[0107] Step 702, the original text obtained by the mobile terminal human-computer interaction interface is "Who are the doctors who treat uremia in the Nephrology Department of XX Hospital?";

[0108] Step 704, extracting segmentation words and keywords; "place name: XX Hospital", "department name: Nephrology Department", "diagnosis: uremia", "job title: doctor", "question word: which", question type matching; "question type: inquirer, question word: which, expected answer type: name";

[0109] Step 706, constructing a knowledge graph; "Department-Diagnosis, Nephrology Department-Uremia", "Doctor-Diagnosis, Doctor-Uremia", "Institution-Department, XX Hospital-Nephrology Department";

[0110] Step 708, determining that the knowledge graph subgraph is "nephrology";

[0111] Step 710, perform keyword search in the nephrology knowledge graph subgraph, determine that the keyword is uremia, the scope is XX Hospital, and the question type is the inquirer (doctor);

[0112] Step 712, screening the candidate answers in the candidate answer set according to the label of each candidate answer, and / or the text feature and non-text feature;

[0113] Step 714, output the best answer on the mobile terminal human-computer interaction interface: Doctor A. To support the personalized needs of users, multiple candidate answers can be output according to the order of the candidate answers, such as: Doctor A, Doctor B, Doctor C.

[0114] Step 702 and step 714 can be input and output to each other. Based on the iteration of steps 704 to 708, recommended questions are generated and pushed to the mobile terminal, and the question and answer recommendation process is restarted after selection is made using the human-computer interaction interface.

[0115] In this embodiment, statistical analysis, knowledge graph construction and feature selection can be combined to automatically recommend medical question-and-answer scenarios under natural language, and the candidate answers can be constrained by combining clinical expert experience and medical features to ensure the accuracy and professionalism of the recommended answers.

[0116] The second embodiment of the present invention provides a medical question and answer recommendation system 800, such as Figure 8 As shown, it includes a memory 802, a processor 804, and a computer program stored in the memory 802 and executable on the processor 804. When the processor 804 executes the computer program, it realizes: obtaining medical problems; preprocessing the medical problems and determining multiple keywords; constructing a knowledge graph triple set based on multiple keywords; querying a candidate answer set for the medical problem in the existing knowledge graph based on the knowledge graph triple set; determining the answer quality of each candidate answer in the candidate answer set to extract the best answer and push it. Among them, the purpose of obtaining medical problem data is divided into two parts. The first is to first obtain a large amount of medical problem data, perform word segmentation entity recognition, etc. to construct an existing knowledge graph and answer matching, etc., which is the training process; the second is to obtain a medical problem through a mobile terminal, extract keywords, and then enter the recommended final answer.

[0117] The medical question and answer recommendation system 800 provided in this embodiment can combine natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level and understand the medical questions raised by users. At the same time, it combines statistical analysis, knowledge graph construction and feature selection, and uses human-computer interaction to automatically recommend medical question and answer scenarios under natural language. While ensuring the accuracy and professionalism of the recommended answers and meeting the needs of users for obtaining health and medical related knowledge, it improves the utilization rate of medical information resources and information query efficiency, saving redundant working time for medical workers.

[0118] In one embodiment of the present invention, preferably, when the processor 804 executes the computer program, it specifically implements the step of preprocessing the medical question to obtain multiple keywords, including: performing word segmentation and entity recognition on the medical question according to the sentence pattern of the medical question and a preset medical corpus to obtain multiple keywords.

[0119] In this embodiment, according to the sentence pattern of the medical question and the preset medical corpus, the inverse maximum matching algorithm and the multi-layer conditional random field are used to perform word segmentation and entity recognition on the user's medical question to determine the keywords. The medical question and answer recommendation method provided by the present invention can combine natural language processing technology and deep learning algorithms to deeply mine medical free text data from a semantic level, thereby improving information query efficiency.

[0120] In a specific embodiment, the word segmentation tool is optimized according to the characteristics of the medical field, and the medical corpus data structure is defined based on the foreign medical ontology structure.

[0121] In one embodiment of the present invention, preferably, when the processor 804 executes the computer program, it further specifically implements: performing statistics on the multiple keywords, and weighting each of the multiple keywords based on the statistical features.

[0122] In this embodiment, statistics are taken on multiple keywords, and each of the multiple keywords is weighted based on statistical features to obtain keywords with higher relevance, ensure the accuracy of semantic understanding in a specific field, and improve the utilization rate of medical information resources and information query efficiency.

[0123] In a specific embodiment, the statistical features include any one of the following or a combination thereof: word frequency, position, and co-occurrence frequency.

[0124] In one embodiment of the present invention, preferably, when the processor 804 executes the computer program, it specifically implements the step of constructing a knowledge graph triple set based on multiple keywords, including: extracting medical entities and entity relationships from multiple keywords, performing synonym conversion, and constructing a knowledge graph triple set in the form of triple storage.

[0125] In this embodiment, in the field of medical question and answer, the authority of experts is very important. Therefore, methods such as conditional random fields are used to extract medical entities and entity relationships from multiple keywords, and synonym conversion is performed. A knowledge graph triple set is constructed in the form of triple storage. By constructing the knowledge graph triple set, keywords are integrated and a resource description framework (RDF) is defined to facilitate understanding of medical questions raised by users, thereby improving the utilization rate of medical information resources and information query efficiency.

[0126] In a specific embodiment, the triple set includes: resource, attribute type and attribute value.

[0127] In one embodiment of the present invention, preferably, the processor 804 executes the computer program to specifically implement the step of querying a set of candidate answers to medical questions in an existing knowledge graph based on a knowledge graph triple set, including: determining a knowledge graph subgraph based on the weight of each keyword; matching a knowledge graph path based on the knowledge graph triple set; searching in the knowledge graph subgraph based on the knowledge graph path to query a set of candidate answers to the medical question.

[0128] In this embodiment, considering the complexity of medical knowledge and the large differences in characteristics of various diseases, the knowledge graph subgraph is determined according to the weight of each keyword, the query scope is narrowed, and the knowledge graph path is matched according to the knowledge graph triple set. The medical entity and entity relationship are combined to constrain the retrieval path. The keyword matching technology is used to search in the knowledge graph subgraph according to the knowledge graph path, and the words and phrases closest to the medical problem are selected to form a set of candidate answers. While ensuring the accuracy and professionalism of the recommended answers, the utilization rate of medical information resources and the efficiency of information query are improved, saving redundant working time for medical workers.

[0129] In one embodiment of the present invention, preferably, when the processor 804 executes the computer program, it specifically implements the steps of extracting the best answer from the candidate answer set and pushing it, including: screening the candidate answers in the candidate answer set according to the label of each candidate answer, and / or text features and non-text features; sorting the screened one or more candidate answers to determine the best answer.

[0130] In this embodiment, the label of each candidate answer is analyzed, and higher-quality answers are further screened based on the label and / or text features and non-text features. The maximum entropy method is used to predict the answer quality, the candidate answers are ranked, and the best answer is selected, thereby ensuring the accuracy and professionalism of the recommended answers and meeting the user's needs for obtaining health and medical related knowledge.

[0131] The medical question and answer recommendation method and system provided by the embodiment of the present invention include three main parts: question understanding, information retrieval, and answer extraction. The question understanding part is to pre-process and understand keywords of questions raised by users in natural language, and obtain the focus of the user's query sentence and the type of user's question; the information retrieval part is to obtain the documents where the answers may be located through search engines and knowledge graphs, and sort the document recommendations; the answer extraction part is to further process the candidate documents obtained by information retrieval, select the best answer and return it to the user. The application of the medical question and answer recommendation system fully improves the utilization rate of medical information resources and saves redundant working time for medical workers.

[0132] The third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned medical question and answer recommendation methods are implemented. Therefore, the computer-readable storage medium includes all the beneficial effects of the medical question and answer recommendation method provided by any of the above-mentioned embodiments.

[0133] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance unless otherwise clearly specified and limited; the terms "connection", "installation", "fixation" and the like should be understood in a broad sense, for example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0134] In the description of this specification, the description of the terms "one embodiment", "some embodiments", "specific embodiments", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0135] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A medical question and answer recommendation method, characterized in that: include: Get medical questions; Preprocessing the medical problem to determine multiple keywords; Construct a knowledge graph triple set based on the multiple keywords to obtain the focus of the user query and the type of the user question; wherein the triple set includes: resources, attribute types and attribute values; According to the knowledge graph triple set, querying the existing knowledge graph for a set of candidate answers to the medical question; Extracting the best answer from the candidate answer set and pushing it; The step of pre-processing the medical problem to obtain a plurality of keywords includes: According to the sentence pattern of the medical question and the preset medical corpus, the medical question is segmented and entity recognized by using the reverse maximum matching algorithm and the multi-layer conditional random field to obtain the multiple keywords; Also includes: Performing statistics on the multiple keywords, and weighting each of the multiple keywords based on statistical features, so as to obtain keywords with high relevance and ensure the accuracy of semantic understanding in a specific field; the statistical features include any one of the following or a combination thereof: word frequency, position, and co-occurrence frequency; The step of constructing a knowledge graph triple set according to the multiple keywords includes: Extracting medical entities and entity relationships from the multiple keywords, performing synonym conversion, constructing the knowledge graph triple set in the form of triple storage, integrating the keywords by constructing the knowledge graph triple set, and defining a resource description framework; The step of searching for a set of candidate answers to the medical question in an existing knowledge graph according to the knowledge graph triple set includes: Determine a knowledge graph subgraph according to the weight of each of the keywords; Match the knowledge graph path according to the knowledge graph triple set; constrain the retrieval path by combining medical entities and entity relationships; According to the knowledge graph path, a search is performed in the knowledge graph subgraph using keyword matching technology to query a set of candidate answers to the medical question.

2. The medical question and answer recommendation method according to claim 1, characterized in that: The step of extracting the best answer from the candidate answer set and pushing it includes: Screening candidate answers in the candidate answer set according to the label of each candidate answer, and / or text features and non-text features; The screened one or more candidate answers are ranked to determine the best answer.

3. A medical question-answer recommendation system, characterized in that: The invention comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following is achieved: Get medical questions; Preprocessing the medical problem to determine multiple keywords; Construct a knowledge graph triple set based on the multiple keywords to obtain the focus of the user query and the type of the user question; wherein the triple set includes: resources, attribute types and attribute values; According to the knowledge graph triple set, querying the existing knowledge graph for a set of candidate answers to the medical question; Determine the answer quality of each candidate answer in the candidate answer set to extract the best answer and push it; When the processor executes the computer program, the step of preprocessing the medical problem to obtain a plurality of keywords includes: According to the sentence pattern of the medical question and the preset medical corpus, the medical question is segmented and entity recognized by using the reverse maximum matching algorithm and the multi-layer conditional random field to obtain the multiple keywords; When the processor executes the computer program, it further specifically implements: Performing statistics on the multiple keywords, and weighting each of the multiple keywords based on statistical features, so as to obtain keywords with high relevance and ensure the accuracy of semantic understanding in a specific field; the statistical features include any one of the following or a combination thereof: word frequency, position, and co-occurrence frequency; When the processor executes the computer program, the step of constructing a knowledge graph triple set according to the multiple keywords includes: Extracting medical entities and entity relationships from the multiple keywords, performing synonym conversion, constructing the knowledge graph triple set in the form of triple storage, integrating the keywords by constructing the knowledge graph triple set, and defining a resource description framework; When the processor executes the computer program, the step of searching for a set of candidate answers to the medical question in an existing knowledge graph according to the knowledge graph triple set includes: Determine a knowledge graph subgraph according to the weight of each of the keywords; Match the knowledge graph path according to the knowledge graph triple set; constrain the retrieval path by combining medical entities and entity relationships; According to the knowledge graph path, a search is performed in the knowledge graph subgraph using keyword matching technology to query a set of candidate answers to the medical question.

4. The medical question and answer recommendation system according to claim 3, characterized in that: When the processor executes the computer program, the steps of extracting the best answer from the candidate answer set and pushing the best answer include: Screening candidate answers in the candidate answer set according to the label of each candidate answer, and / or text features and non-text features; The screened one or more candidate answers are ranked to determine the best answer.

5. A computer-readable storage medium, characterized in that: A computer program is stored thereon, characterized in that when the computer program is executed by a processor, the steps of the medical question and answer recommendation method as claimed in claim 1 or 2 are implemented.

Images