Honey or Poison? Solving the Trigger Curse in Few-shot Event Detection via Causal Intervention
This work addresses a specific bottleneck in event detection for NLP researchers, offering a novel solution but likely incremental in the broader context of causal methods.
The paper tackled the trigger curse problem in few-shot event detection, where overfitting or underfitting triggers harms generalization or performance, by using causal intervention to adjust for the trigger as a confounder, resulting in significant improvements on ACE05, MAVEN, and KBP17 datasets.
Event detection has long been troubled by the \emph{trigger curse}: overfitting the trigger will harm the generalization ability while underfitting it will hurt the detection performance. This problem is even more severe in few-shot scenario. In this paper, we identify and solve the trigger 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 previous FSED methods much easier to overfit triggers. To resolve this problem, we propose to intervene on the context via backdoor adjustment during training. Experiments show that our method significantly improves the FSED on ACE05, MAVEN and KBP17 datasets.