Biomedical text representation method for modeling global and local context interaction

Abstract

The invention discloses a biomedical text representation method for modeling global and local context interaction, and belongs to the field of biomedical text representation methods. According to themethod, a given biomedical text is a sequence (S1,..., Si,..., SL) consisting of L sentences, wherein each sentence Si is expressed as a sequence composed of the words in the sentence; each sentence Si in the given text can be expressed as a matrix Xi through an input module, and the initial expression, obtained by the input module, of the biomedical text is input into an expression learning module for further text modeling, wherein each word is used as a node, a syntax dependence tree obtained by an open source tool Stanford CoreNLP is used as a topological structure between the nodes, and modeling is performed on local context information in Si by applying two layers of GCNs; a hypergraph concept is introduced to aggregate the local context information to obtain a representation that a corresponding node in the hypergraph contains global context information; and finally information interaction of local and global contexts is modeled to learn rich representations of related concepts in each sentence.

Description

technical field [0001] The invention relates to a biomedical text representation method, in particular to a biomedical text representation method for modeling global and local context interaction, and belongs to the technical field of biomedical text representation methods. Background technique [0002] In recent years, the field of biomedicine has developed vigorously, and biomedical literature has shown an explosive growth trend. How to quickly and accurately obtain target information from a large amount of biomedical text data is a topic with application prospects and research significance. Effective modeling of biomedical text is the basis for effective information extraction. Existing text modeling methods can be roughly divided into three categories: (1) Traditional text modeling methods, using feature selection or feature extraction methods to obtain text features, and on this basis, apply classic classification or clustering algorithms for classification and cluster ...

AI Technical Summary

Problems solved by technology

Therefore, direct application of existing text modeling methods cannot obtain ideal text representation, which will affect the effect of downstream bioinformatics tasks. Therefore, a biomedical text representation method for modeling global and local context interactions is designed to optimize above question

Images