Construction And Application Of Medical Knowledge Graph Based On Cross-sentence Relation Extraction

In recent years,the Internet field has shown explosive development.There are thousands of web pages data constantly updated every day,and hundreds of knowledge in each web page text waiting for people to explore.In order to use this information effectively,domestic and foreign researchers have done a lot of research work on it.Among them,knowledge graph construction is one of the most important and popular research directions.This article analyzes and researches how to realize a complete medical knowledge graph and how to realize medical knowledge question and answer based on the medical knowledge graph.The main innovations and improvements of this article are as follows:(1)This article discovers and uses the medical classification hierarchical directory features in the internet web documents to make the construction of medical ontology more accurate and efficient;(2)This article uses a self-designed crawling algorithm obtain a large amount of the internet medical text data,and perform a series of processing on it to obtain a brand-new cross-sentence relation extraction and annotation data-set;(3)This article introduces the method of cross-sentence relation extraction in the process of constructing medical knowledge graph.This paper analyzes and studies how to realize a complete medical knowledge graph and realize medical knowledge question and answer based on the medical knowledge graph.Through the research on the construction of knowledge graph for the medical field,it realizes the extraction and application of a large number of medical knowledge data existing in the Internet.Firstly,this article uses Scrapy and Jsoup Api to crawl a large amount of unstructured medical text data from websites such as 39 health.com and Dingxiangyuan.Secondly,by analyzing these data and combining medical hierarchical catalog information,the medical ontology modeling is completed;Then,medical knowledge is extracted from these unstructured text data through entity extraction and relationship classification methods.Among them,the entity extraction is completed through the BERT-Bi LSTM-CRF model,and the cross-sentence entity is completed through the GA-LSTM model.According to the triad of knowledge obtained by the above extraction,a hierarchical medical knowledge graph is constructed and perfected;finally,the medical knowledge is stored correspondingly by using the Neo4 j graph database,and based on the medical knowledge graph design and Realize such as medical entity extraction.Multiple functions such as medical entity query,medical relationship extraction,medical relationship query,medical knowledge question and answer,medical knowledge extraction and import,fully realize the intuitive display and convenient use of medical knowledge.