Joint entity relation extraction of Chinese electronic medical record based on graph convolutional neural network and words for relationship discovery

Abstract

Considering the characteristics of Chinese medical records, such as many short sentences, weak long-distance relations, and serious relation overlap, a joint entity relation extraction method based on relation discovery word and graph convolution is proposed. This method is based on the relationship discovery word set. Through the word attention mechanism, different weights are assigned to the context description words of the relationship to achieve semantic enhancement. Based on the text-embedded representation obtained from the pre-trained language model, a feature fusion gating mechanism was designed to realize the organic fusion of the output features of the bidirectional short and long-duration memory network and the feature of relation discovery words. The weighted dependency graph, which is encoded by a graph convolutional neural network and pruned by an attentional mechanism, is used to obtain local features of sentence grammatical structure. The proposed method was applied to CMeIE and CMRT-RE data sets, and compared with other mainstream methods, the effectiveness of the proposed method in joint extraction of medical entity-relationship was verified experimentally.