A Two-Step Approach for Implicit Event Argument Detection

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7479–7485 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 7479 A Two-Step Approach for Implicit Event Argument Detection Zhisong Zhang, Xiang Kong, Zhengzhong Liu, Xuezhe Ma, Eduard Hovy Language Technologies Institute, Carnegie Mellon University {zhisongz,xiangk,liu,xuezhem,hovy}@cs.cmu.edu Abstract In this work, we explore the implicit event ar- gument detection task, which studies event ar- guments beyond sentence boundaries. The ad- dition of cross-sentence argument candidates imposes great challenges for modeling. To reduce the number of candidates, we adopt a two-step approach, decomposing the problem into two sub-problems: argument head-word detection and head-to-span expansion. Evalu- ated on the recent RAMS dataset (Ebner et al., 2020), our model achieves overall better per- formance than a strong sequence labeling base- line. We further provide detailed error analy- sis, presenting where the model mainly makes errors and indicating directions for future im- provements. It remains a challenge to de- tect implicit arguments, calling for more future work of document-level modeling for this task. 1 Introduction Event argument detection is a key component in the task of event extraction. It resembles semantic role labeling (SRL) in that the main target is to find argument spans to fill the roles of event frames. However, event arguments can go beyond sentence boundaries: there can be non-local or implicit argu- ments at the document level. Figure 1 shows such an example: for the purchase event, which is trig- gered by the word “bought”, its money argument appears in the previous sentence. Implicit arguments have been under-explored in event extraction. Most of previous systems (Li et al., 2013; Chen et al., 2015; Nguyen et al., 2016; Wang et al., 2019) only consider local arguments in the same sentence of the event trigger. While incor- porating implicit arguments requires corresponding annotations, few exists in most of the widely used event datasets, like ACE2005 (LDC, 2005; Walker et al., 2006) and RichERE (LDC, 2015). There are several annotation efforts for implicit arguments (a) The new computer cost 3000 dollars, while the old one cost 1000 dollars. Nevertheless, he still bought the more expensive one. (b) The new computer cost 3000 dollars, while the old one cost 1000 dollars. Therefore, he bought the cheaper one. Tokens: ... cost 3000 dolloars , ... ... still bought the more ... Indicators: ... 0 0 0 0 ... ... 1 0 1 1 ... BERT Encoder ... ... ... ... Biaffine Role-Scorer Head-to-Span Expander left right arg trigger BERT Reprs. Decoders Figure 1: Examples of implicit arguments and model illustration. The bold text indicates the trigger word for the purchase event, while the underlined text indi- cates its non-local “money” argument in the previous sentence. Our model first detects the head-word “dol- lars”, and then expands it to the whole span. in SRL, including G&C (Gerber and Chai, 2010, 2012), SemEval-2010 (Ruppenhofer et al., 2009, 2010), and 80Days (Feizabadi and Pad´o, 2014). Yet most are performed with different ontologies such as Nombank …

Content truncated to 500 words