Research And Implementation Of Medical Commercial Knowledge Graph Construction Based On Multi-source

Knowledge graph is a technology that integrates multi-source heterogeneous knowledge into a unified network structure model and performs knowledge inference and mining.Since the development of knowledge graph,it is not limited to the field of general knowledge.The application of knowledge graph in the professional field can carry out the structured sorting and deep analysis of the framework structure of the complex model of professional knowledge,accurately and clearly show the internal connection of professional knowledge.Knowledge reasoning divergence and assisting decision-making play an important role.This paper studies and constructs a business knowledge graph in the medical field,and its main contributions are as follows:(1)Design definition of atlas ontology layer.Based on the ontology theoretical knowledge,combined with the analysis of the original data model,extract the main ontology class and the relationship between ontologies,define the entities,entity attributes,relationships and relationship attributes in the conceptual layer of the business knowledge graph,and use RDFS language to describe(2)Acquire and process data from multiple data sources.Design a crawler system to collect a large amount of text-type data and semi-structured data from a distributed information publishing platform for data preparation for knowledge graph construction;(3)Propose a named entity recognition system based on Bi-LSTM-CRF model.Use word2 vec for vectorized representation of text,Bi-LSTM model for feature extraction,and conditional random field model for sequence annotation.In the comparative experiment,the F1 value of the model name entity,person name entity,and product name entity identification reached more than 85%,and the F1 value of the organization name entity identification reached 91.97%;(4)An algorithm for event extraction and summary generation based on machine learning is proposed.The doc2 vec model is used for document-level vectorized representation,the K-Means algorithm is used for event document clustering,and the algorithm based on keyword coverage is used for document abstract extraction.In this paper,the ROUGE-L value of the abstract extraction algorithm reaches 0.47 in the experiment.(5)Construction and verification of medical business knowledge map.On the basis of the definition of the pattern layer and knowledge extraction,the actual construction of the map is carried out.Based on the medical business knowledge map constructed in this paper,users can quickly and comprehensively understand the relevant information of enterprises and hospitals,tap deep-level business relationships,and provide technical support for business decisions.