Document-Level Event Argument Extraction by Conditional Generation

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 894–908 June 6–11, 2021. ©2021 Association for Computational Linguistics 894 Document-Level Event Argument Extraction by Conditional Generation Sha Li and Heng Ji and Jiawei Han University of Illinois at Urbana-Champaign, IL, USA {shal2, hengji, hanj}@illinois.edu Abstract Event extraction has long been treated as a sentence-level task in the IE community. We argue that this setting does not match hu- man information seeking behavior and leads to incomplete and uninformative extraction results. We propose a document-level neu- ral event argument extraction model by for- mulating the task as conditional generation following event templates. We also com- pile a new document-level event extraction benchmark dataset WIKIEVENTS which in- cludes complete event and coreference anno- tation. On the task of argument extraction, we achieve an absolute gain of 7.6% F1 and 5.7% F1 over the next best model on the RAMS and WIKIEVENTS datasets respectively. On the more challenging task of informative ar- gument extraction, which requires implicit coreference reasoning, we achieve a 9.3% F1 gain over the best baseline. To demonstrate the portability of our model, we also create the first end-to-end zero-shot event extraction framework and achieve 97% of fully super- vised model’s trigger extraction performance and 82% of the argument extraction perfor- mance given only access to 10 out of the 33 types on ACE. 1 1 Introduction By converting a large amount of unstructured text into trigger-argument structures, event extraction models provide unique value in assisting us process volumes of documents to form insights. While real-world events are often described throughout a news document (or even span multiple documents), the scope of operation for existing event extraction models have long been limited to the sentence level. Early work on event extraction originally posed the task as document level role filling (Grishman and Sundheim, 1996) on a set of narrow scenarios 1The programs, data and resources are publicly avail- able for research purpose at https://github.com/ raspberryice/gen-arg. Prosecutors say he drove the truck to Geary Lake in Kansas, that 4,000 pounds of ammonium nitrate laced with nitromethane were loaded into the truck there, and that it was driven to Oklahoma City and detonated. Elliott testified that on April 15, McVeigh came into the body shop and reserved the truck, to be picked up at 4pm two days later. Elliott said that McVeigh gave him the $280.32 in exact change after declining to pay an additional amount for insurance. Figure 1: Two examples of cross-sentence infer- ence for argument extraction from our WIKIEVENTS dataset. The PaymentBarter argument of the Transac- tion.ExchangeBuySell event triggered by “reserved” in the first sentence can only be found in the next sentence. The Attack.ExplodeDetonate event triggered by “deto- nated” in the third sentence has an uninformative ar- gument “he”, which needs to be resolved to the name mention “McVeigh” in the previous sentences. and evaluated on small datasets. The release of ACE2, a large scale dataset with …

Content truncated to 500 words