Empower Chinese event detection with improved atrous convolution neural networks
- PDF / 4,043,452 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 63 Downloads / 214 Views
(0123456789().,-volV)(0123456789(). ,- volV)
REVIEW
Empower Chinese event detection with improved atrous convolution neural networks Zhihong Wang1 • Yi Guo1,2,3
•
Jiahui Wang1
Received: 14 January 2020 / Accepted: 9 September 2020 Ó Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract Event Detection (ED) is an essential and challenging task in Information Extraction (IE). Recent advances in neural networks make it possible to build reliable models without complicated feature engineering. Neural network-based models commonly regard event detection as a char-wise or word-wise labeling task, which suffers from the problems of longdistance information miss-capturing, discontinuous labeling errors, etc., between characters/words and event triggers, especially in languages without natural word delimiters such as Chinese. In this paper, we propose a novel multi-layer Residual and Gated-Based Atrous Convolution Neural Network (RG-ACNN) framework, which attempts to alleviate above-mentioned problems. Specifically, the ACNN is introduced in our model to expand the receptive field to obtain multi-scale context information to capture dependencies between long-distance information. While gated and residual mechanisms are both imported to ACNN to improve our networks’ capability of the information filtering and aggregation. Besides, RG-ACNN performs event detection in a char-wise paradigm, where a novel ‘‘head-tail dual-pointer’’ labeled strategy is used to overcome the incomplete continuous labeling problem. Experiments on the ACE2005-CN and several standard benchmark datasets show that RG-ACNN significantly outperforms state-of-the-art (SOTA) methods. Keywords Atrous convolution networks Gated mechanism Residual mechanism Head-tail dual-pointer Event detection
1 Introduction Automatic event detection is a crucial and challenging subtask of event extraction. Given a text document, an event detection system should predict event triggers with specific types and sub-types. For example, from the sentence ‘‘Shanghai Public Security Bureau (SPSB) apprehend them & Yi Guo [email protected] Zhihong Wang [email protected] Jiahui Wang [email protected] 1
East China University of Science and Technlogy, Shanghai, China
2
National Engineering Laboratory for Big Data Distribution and Exchange Technologies, Shanghai, China
3
Shanghai Engineering Research Center of Big Data & Internet Audience, Shanghai, China
all with an opportunity, while the unpopular Guangnan Yang was also deported and sent back to Macau.’’ An event detection model should detect a ‘‘Justice/Arrest-Jail’’ event triggered by ‘‘apprehend,’’ and a ‘‘Justice/Extradite’’ event triggered by ‘‘back to.’’ Recently, event detection is generally transformed into a sequence labeling and classification problem. For each event sentence, sequence labeling and classification methods seek to assign a reasonable categorical label (trigger word/event type) to each member of the event sentence sequence, including word-wise event detection methods and char-
Data Loading...
