Joint Event Extraction via Recurrent Neural Networks

Proceedings of NAACL-HLT 2016, pages 300–309, San Diego, California, June 12-17, 2016. c⃝2016 Association for Computational Linguistics Joint Event Extraction via Recurrent Neural Networks Thien Huu Nguyen, Kyunghyun Cho and Ralph Grishman Computer Science Department, New York University, New York, NY 10003, USA thien@cs.nyu.edu,kyunghyun.cho@nyu.edu,grishman@cs.nyu.edu Abstract Event extraction is a particularly challenging problem in information extraction. The state- of-the-art models for this problem have ei- ther applied convolutional neural networks in a pipelined framework (Chen et al., 2015) or followed the joint architecture via structured prediction with rich local and global features (Li et al., 2013). The former is able to learn hidden feature representations automatically from data based on the continuous and gen- eralized representations of words. The latter, on the other hand, is capable of mitigating the error propagation problem of the pipelined ap- proach and exploiting the inter-dependencies between event triggers and argument roles via discrete structures. In this work, we propose to do event extraction in a joint framework with bidirectional recurrent neural networks, thereby benefiting from the advantages of the two models as well as addressing issues inher- ent in the existing approaches. We systemati- cally investigate different memory features for the joint model and demonstrate that the pro- posed model achieves the state-of-the-art per- formance on the ACE 2005 dataset. 1 Introduction We address the problem of event extraction (EE): identifying event triggers of specified types and their arguments in text. Triggers are often single verbs or normalizations that evoke some events of interest while arguments are the entities participating into such events. This is an important and challeng- ing task of information extraction in natural lan- guage processing (NLP), as the same event might be present in various expressions, and an expression might expresses different events in different con- texts. There are two main approaches to EE: (i) the joint approach that predicts event triggers and arguments for sentences simultaneously as a structured predic- tion problem, and (ii) the pipelined approach that first performs trigger prediction and then identifies arguments in separate stages. The most successful joint system for EE (Li et al., 2013) is based on the structured perceptron al- gorithm with a large set of local and global fea- tures1. These features are designed to capture the discrete structures that are intuitively helpful for EE using the NLP toolkits (e.g., part of speech tags, de- pendency and constituent tags). The advantages of such a joint system are twofold: (i) mitigating the er- ror propagation from the upstream component (trig- ger identification) to the downstream classifier (ar- gument identification), and (ii) benefiting from the the inter-dependencies among event triggers and ar- gument roles via global features. For example, con- sider the following sentence (taken from Li et al. (2013)) in the ACE 2005 dataset: In Baghdad, a cameraman died when an Ameri- can tank fired on the Palestine hotel. In this sentence, died and fired are the event trig- gers for the events of types Die and Attack, respec- tively. In the pipelined …

Content truncated to 500 words