Weighted-Dependency with Attention-Based Graph Convolutional Network for Relation Extraction

Due to the complexity of natural language, the current relation extraction methods can no longer meet people's requirements. Dependency trees have been proved to be able to capture the long-distance relation between the target entity pairs, this study mainly focuses on how to prune the dependency tree more effectively, and improve the model's performance by retaining the original dependency structure of the sentence and dependency relation weighting without man-made noise. We propose a new weighted-dependency with attention-based graph convolutional network (WAGCN), which changed the dependency matrix from a 0-1 matrix to a weighted matrix by weighting each dependency class to enable the dependency matrix to fully express the syntactic dependency information in the sentence. Moreover, we improve the connection structure of the attention mechanism, densely connected network and GCN to extract important information by applying the attention mechanism directly to the text, and compute it in parallel with the dependency matrix GCN module. We conducted experiments on three popular datasets, WAGCN is superior to baseline models, such as C-AGGCN and LST-AGCN, in most experiment metrics.