Event factuality identification aims at determining the factual nature of events, and plays an important role in NLP. This paper proposes a two-step framework for identifying the factuality of events in raw texts. Firstly, it extracts various basic factors related with factuality of events. Then, it utilizes a hybrid neural network model which combines Bidirectional Long Short-Term Memory (BiLSTM) networks and Convolutional Neural Networks (CNN) for event factuality identification, and considers lexical and syntactic features. The experimental results on FactBank show that the proposed neural network model significantly outperforms several state-of-the-art baselines, particularly on speculative and negative events.
CITATION STYLE
Qian, Z., Li, P., Zhou, G., & Zhu, Q. (2018). Event factuality identification via hybrid neural networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11305 LNCS, pp. 335–347). Springer Verlag. https://doi.org/10.1007/978-3-030-04221-9_30
Mendeley helps you to discover research relevant for your work.