Event Detection with Neural Networks: A Rigorous Empirical Evaluation
This work addresses the need for systematic comparisons in event detection for natural language processing, but it is incremental as it builds on existing neural approaches.
The paper tackled the problem of event detection and classification from text by rigorously evaluating neural network architectures, finding that their novel GRU-based model with syntactic and temporal attention is competitive with other methods on the ACE2005 dataset.
Detecting events and classifying them into predefined types is an important step in knowledge extraction from natural language texts. While the neural network models have generally led the state-of-the-art, the differences in performance between different architectures have not been rigorously studied. In this paper we present a novel GRU-based model that combines syntactic information along with temporal structure through an attention mechanism. We show that it is competitive with other neural network architectures through empirical evaluations under different random initializations and training-validation-test splits of ACE2005 dataset.