Event Extraction as Machine Reading Comprehension

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 1641–1651, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 1641 Event Extraction as Machine Reading Comprehension Jian Liu1,2,3, Yubo Chen1,2, Kang Liu1,2, Wei Bi4, Xiaojiang Liu4 1 National Laboratory of Pattern Recognition, Institute of Automation Chinese Academy of Sciences, Beijing, 100190, China 2 University of Chinese Academy of Sciences 3 Beijing Jiaotong University, 100044, China 4 Tencent AI Lab, Shenzhen, 518000, China {jian.liu, yubo.chen, kliu}@nlpr.ia.ac.cn {victoriabi, kieranliu}@tencent.com Abstract Event extraction (EE) is a crucial informa- tion extraction task that aims to extract event information in texts. Previous methods for EE typically model it as a classification task, which are data-hungry and suffer from the data scarcity problem. In this paper, we pro- pose a new learning paradigm of EE, by ex- plicitly casting it as a machine reading com- prehension problem (MRC). Our approach in- cludes an unsupervised question generation process, which can transfer event schema into a set of natural questions, followed by a BERT- based question-answering process to retrieve answers as EE results. This learning paradigm enables us to strengthen the reasoning process of EE, by introducing sophisticated models in MRC, and relieve the data scarcity problem, by introducing the large-scale datasets in MRC. The empirical results show that: i) our ap- proach attains state-of-the-art performance by considerable margins over previous methods. ii) Our model is excelled in the data-scarce sce- nario, for example, obtaining 49.8% in F1 for event argument extraction with only 1% data, compared with 2.2% of the previous method. iii) Our model also fits with zero-shot scenar- ios, achieving 37.0% and 16% in F1 on two datasets without using any EE training data. 1 Introduction Event extraction (EE), a crucial information extrac- tion (IE) task, aims to extract event information in texts. For example, in a sentence S1 (shown in Figure 1 (a)), an EE system should recognize an Attack event1, expressed by an event trig- ger stabbed with four event arguments — Sun- day (Role=Time), a protester (Role=Attacker), an officer (Role=Target), and a paper cutter (Role=Instrument). EE is shown to benefit a wide range of applications including knowledge 1According to the ACE event ontology. Figure 1: Comparison of the event extraction task and machine reading comprehension task. base augmentation (Ji and Grishman, 2011), docu- ment summarization, question answering (Berant et al., 2014), and others. In the current study, EE is mostly formulated as a classification problem, aiming to locate and categorize each event trigger/argument (Ahn, 2006; Li et al., 2013; Chen et al., 2015; Nguyen et al., 2016). Despite many advances, classification based methods are data-hungry, which require a great deal of training data to ensure good performance (Chen et al., 2017; Li et al., 2013; Liu et al., 2018a). Moreover, such methods generally cannot deal with new event types never encountered during training time (Huang et al., 2018). In this particular study, we introduce a new learn- ing paradigm for EE, shedding lights on tackling the above problems simultaneously. …

Content truncated to 500 words