Identifying Chinese Event Factuality with Convolutional Neural Networks

Event factuality describes the factual level of the event expressed by event narrator and is one of the deep semantic representations of natural texts. This paper focuses on identifying Chinese event factuality and proposes an effective approach based on CNN (Convolutional Neural Networks). It extracts factual related information from event sentences and then regards them and their transformation as features. Meanwhile, it transfers the features to word vectors to construct a sentence-level word vector map. Finally, it inputs the word vector map to the CNN model to identify event factuality. Experimental results show that our approach achieves a higher performance by using factual features and CNN model, especially the advantage to tackle the imbalanced data distribution problem.