Research And System Implementation Of Entity Relation Extraction Algorithm Based On Text Generation

Entity relationship extraction,as one of the core tasks in the field of natural language processing and information extraction,is to extract factual information,such as specified entities and relations,from unstructured or semi-structured text through relation extraction technology and save it in a structured form.However,most of the currently published researches focus on processing simple scenarios,transforming entity relation extraction into sentence-level classification task,and it is difficult to deal with situations where multiple entities and multiple relations are contained in a sentence that is common in real scenarios.In the medical related fields,the level of health information is constantly improving,and the medical information system is constantly improving,gathering a large number of medical data.It is a key problem to extract structured information from data with different modes for management,sharing and application.However,there are numerous and complex data with different forms in the medical field.It is still difficult to provide structured high-quality knowledge from medical unstructured and semi-structured text to provide data support for subsequent knowledge graph construction.In response to the above problems and challenges,the main contents of this paper are as follows:(1)A multiple entity relation extraction method based on pointer-generator is proposed,which converts the entity relation extraction problem into the text generation task,that is,generating entity words and relationship words as the target text to solve the problem of extracting overlapping multi-relations in unstructured text,and improves the performance of relational triple extraction;(2)On this basis,a multiple entity relation extraction method based on Hierarchical LSTMs structure is proposed,the low-level LSTM is used to predict the relation,and the high-level LSTM is used to identify a pair of entities belonging to the current relation,which makes clear the generating position of entities and relations,and better integrates relation information to guide the generation of corresponding entity labeling sequences,improves the accuracy of entity recognition in joint extraction,and further improves the accuracy of extracting relation triples;(3)Entity relation extraction technology for unstructured and semi-structured text in Chinese medical field,to achieve joint extraction of entity relations based on local text,and automatic extraction of structured knowledge for disease encyclopedia pages,and integrate a set of medical structured knowledge extraction tool,and a set of Chinese medical structured knowledge is output.Based on the above models and tools,in view of the lack of open-source entity relation extraction platform in the current market,a web-based medical entity relation extraction service platform is established,which realizes the whole process engineering service platform from data collection,knowledge extraction,and data visualization,and provides technical services for researchers and related users.The system is divided into data collection module,knowledge extraction module,knowledge storage module and visualization module.