Honey or Poison? Solving the Trigger Curse in Few-shot Event Detection via Causal Intervention

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 8078–8088 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 8078 Honey or Poison? Solving the Trigger Curse in Few-shot Event Detection via Causal Intervention Jiawei Chen1,3, Hongyu Lin1,∗, Xianpei Han1,2,∗, Le Sun1,2, 1Chinese Information Processing Laboratory 2State Key Laboratory of Computer Science Institute of Software, Chinese Academy of Sciences, Beijing, China 3University of Chinese Academy of Sciences, Beijing, China {jiawei2020}@iscas.ac.cn {hongyu,xianpei,sunle}@iscas.ac.cn Abstract Event detection has long been troubled by the trigger curse: overfitting the trigger will harm the generalization ability while underfitting it will hurt the detection performance. This prob- lem is even more severe in few-shot scenario. In this paper, we identify and solve the trig- ger curse problem in few-shot event detection (FSED) from a causal view. By formulating FSED with a structural causal model (SCM), we found that the trigger is a confounder of the context and the result, which makes previ- ous FSED methods much easier to overfit trig- gers. To resolve this problem, we propose to intervene on the context via backdoor adjust- ment during training. Experiments show that our method significantly improves the FSED on ACE05, MAVEN and KBP17 datasets. 1 Introduction Event detection (ED) aims to identify and clas- sify event triggers in a sentence, e.g., detecting an Attack event triggered by fire in “They killed by hostile fire in Iraqi”. Recently, supervised ED approaches have achieved promising performance (Chen et al., 2015; Nguyen and Grishman, 2015; Nguyen et al., 2016; Lin et al., 2018, 2019b,a; Du and Cardie, 2020; Liu et al., 2020a; Lu et al., 2021), but when adapting to new event types and domains, a large number of manually annotated event data is required which is expensive. By contrast, few- shot event detection (FSED) aims to build effective event detectors that are able to detect new events from instances (query) with a few labeled instances (support set). Due to their ability to classify novel types, many few-shot algorithms have been used in FSED, e.g., metric-based methods like Prototypical Network (Lai et al., 2020; Deng et al., 2020; Cong et al., 2021). Unfortunately, there has long been a “trigger curse” which troubles the learning of event detec- ∗Corresponding authors. An Attack event in Iraqi. fire They were killed by hostile [MASK] in Iraqi. They were killed by hostile fire in Iraqi. 1 or 0 Query E T C S Y E C T S Y Q (a) Structural Causal Model for the data distribution of FSED An Attack event in Iraqi. forces, fire, attack, ... They were killed by hostile [MASK] in Iraqi. They were killed by hostile [T] in Iraqi. 1 or 0 Query E T C S Y E C T S Y Q (b) The data distribution of FSED after causal intervention Figure 1: Illustration of the causal intervention strategy proposed in this paper. The graph includes the event E, the trigger set T, the context set C, the support instance S, the prediction Y …

Content truncated to 500 words