BAE: BERT-based Adversarial Examples for Text Classification

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 671–683, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 671 Event Extraction by Answering (Almost) Natural Questions Xinya Du and Claire Cardie Department of Computer Science Cornell University Ithaca, NY, USA {xdu, cardie}@cs.cornell.edu Abstract The problem of event extraction requires de- tecting the event trigger and extracting its corresponding arguments. Existing work in event argument extraction typically relies heavily on entity recognition as a preprocess- ing/concurrent step, causing the well-known problem of error propagation. To avoid this issue, we introduce a new paradigm for event extraction by formulating it as a question an- swering (QA) task that extracts the event ar- guments in an end-to-end manner. Empirical results demonstrate that our framework outper- forms prior methods substantially; in addition, it is capable of extracting event arguments for roles not seen at training time (i.e., in a zero- shot learning setting).1 1 Introduction Event extraction is a long-studied and challeng- ing task in Information Extraction (IE) (Sund- heim, 1992; Grishman and Sundheim, 1996; Riloff, 1996). Its goal is to extract structured information — “what is happening” and the persons/objects that are involved — from unstructured text. The task is il- lustrated via an example in Figure 1, which depicts an ownership transfer event (the event type), trig- gered by the word “sale" (the event trigger) and ac- companied by its extracted arguments — text spans denoting entities that fill a set of (semantic) roles as- sociated with the event type (e.g., BUYER, SELLER and ARTIFACT for ownership transfer events). Recent successful approaches to event extraction have benefited from dense features extracted by neural models (Chen et al., 2015; Nguyen et al., 2016; Liu et al., 2018) as well as contextualized lexical representations from pretrained language models (Zhang et al., 2019b; Wadden et al., 2019). 1Our code and question templates for the work are open sourced at https://github.com/xinyadu/eeqa for reproduction purpose. Input: As part of the 11-billion-dollar sale of USA Interactive's film and television operations to the French company and its parent company in December 2001, USA Interactive received 2.5 billion dollars in preferred shares in Vivendi Universal Entertainment. Event type Transaction- Transfer-Ownership Trigger “sale” Args. Buyer “French company”, “parent company” Seller “USA Interactive” Artifact “operations” Place - Extracted Event: Figure 1: Event extraction example from the ACE 2005 corpus (Doddington et al., 2004). The approaches, however, exhibit two key weak- nesses. First, they rely heavily on entity informa- tion for argument extraction. In particular, event argument extraction generally consists of two steps – first identifying entities and their general semantic class with trained models (Wadden et al., 2019) or a parser (Sha et al., 2018), then assigning argu- ment roles (or no role) to each entity. Although joint models (Yang and Mitchell, 2016; Nguyen and Nguyen, 2019; Zhang et al., 2019a; Lin et al., 2020) have been proposed to mitigate this issue, er- ror propagation (Li et al., 2013) still occurs during event argument extraction. A second weakness …

Content truncated to 500 words