BERTering RAMS: What and How Much does BERT Already Know About Event Arguments? -- A Study on the RAMS Dataset

TL;DR

This study investigates BERT's inherent ability to identify event arguments in the RAMS dataset without training, revealing modest initial performance that can be significantly improved with minimal supervision, and explores cross-sentence argument detection and lexical biases.

Contribution

The paper demonstrates BERT's pre-trained knowledge of event arguments, introduces methods to enhance detection with limited supervision, and analyzes cross-sentence and lexical effects on argument identification.

Findings

BERT's attention heads can detect event arguments above chance without training.

Linear combinations of heads improve argument detection accuracy significantly.

Proposed adversarial nonce scheme shows robustness of learned combinations.

Abstract

Using the attention map based probing frame-work from (Clark et al., 2019), we observe that, on the RAMS dataset (Ebner et al., 2020), BERT's attention heads have modest but well above-chance ability to spot event arguments sans any training or domain finetuning, vary-ing from a low of 17.77% for Place to a high of 51.61% for Artifact. Next, we find that linear combinations of these heads, estimated with approx 11% of available total event argument detection supervision, can push performance well-higher for some roles - highest two being Victim (68.29% Accuracy) and Artifact(58.82% Accuracy). Furthermore, we investigate how well our methods do for cross-sentence event arguments. We propose a procedure to isolate "best heads" for cross-sentence argument detection separately of those for intra-sentence arguments. The heads thus estimated have superior cross-sentence performance compared to their jointly estimated equivalents, albeit only under the unrealistic assumption that we already know the argument is present in an-other sentence. Lastly, we seek to isolate to what extent our numbers stem from lexical frequency based associations between gold arguments and roles. We propose NONCE, a scheme to create adversarial test examples by replacing gold arguments with randomly generated "nonce" words. We find that learnt linear combinations are robust to NONCE, though individual best heads can be more sensitive.