Combined Model to Extract Entities and Relations Based on Sharing Parameter

Abstract

This paper uses the depth learning model of sharing parameter to extract entities and relationships. The problems of pipeline model error propagation and ignoring the internal relationship between subtasks, a parameter sharing model is proposed, which uses graph convolution neural network based on syntax to capture the structural information of text. The model combined with the parameter sharing mode will be introduced in detail. The motivation of designing the model, the special labeling strategy, the structure of the model, the experimental setup and the analysis of the experimental results will be introduced respectively. From the experimental results, it can be seen that the hybrid model achieves better results in the public data set.