Event Extraction via Bidirectional Long Short-Term Memory Tensor Neural Networks

Abstract

Traditional approaches to the task of ACE event extraction usually rely on complicated natural language processing (NLP) tools and elaborately designed features. Which suffer from error propagation of the existing tools and take a large amount of human effort. And nearly all of approaches extract each argument of an event separately without considering the interaction between candidate arguments. By contrast, we propose a novel event-extraction method, which aims to automatically extract valuable clues without using complicated NLP tools and predict all arguments of an event simultaneously. In our model, we exploit a context-aware word representation model based on Long Short-Term Memory Networks (LSTM) to capture the semantics of words from plain texts. In addition, we propose a tensor layer to explore the interaction between candidate arguments and predict all arguments simultaneously. The experimental results show that our approach significantly outperforms other state-of-the-art methods.