Entity Relation Extraction Method based on Fusion of Multiple Information and Attention Mechanism
Baosheng Yin, Yifei Sun, Yingxuan Wang · 2020
Entity relation extraction, as the core task and important link in the fields of information extraction, natural language understanding and information retrieval, aims to extract structured entities from a large amount of unstructured text and to extract semantic relations between entities. The entity relation extraction technology based on deep learning has gradually surpassed the traditional method based on features and kernel functions in terms of the depth of feature extraction and the accuracy of model. This paper proposes a multiple model fusion strategy, adds the Convolutional Neural Networks (CNN) on the basis of the original model of Bi-directional Long Short-Term Memory (Bi-LSTM), uses the syntactic structure information of syntactic information and the shortest dependent path separately as input of the experiment, and introduces the attention mechanism to study the dependency relations between the words to capture the internal semantic information of a sentence. The experiment shows that the effect of entity relation extraction based on the fusion of multiple information and attention mechanism is significantly improved than that of a single information model.