Event temporal relation extraction with attention mechanism and graph neural network

Event temporal relation extraction is an important part of natural language processing. Many models are being used in this task with the development deep learning. However, most existing methods cannot accurately obtain degree association between different tokens and events, event-related information be effectively integrated. In paper, we propose event integration model that integrates through multilayer bidirectional long short-term memory (Bi-LSTM) attention mechanism. Although above scheme can improve performance, it still further optimized. To performance previous scheme, a novel relational graph network incorporates edge attributes. approach, first build semantic dependency parsing, considers edges' attributes by using top-k mechanisms to learn hidden contextual representations, finally predict relations. We evaluate proposed on TimeBank-Dense dataset. Compared baselines, Micro-F1 scores obtained our 3.9% 14.5%, respectively.