IDIAPers @ Causal News Corpus 2022: Extracting Cause-Effect-Signal Triplets via Pre-trained Autoregressive Language Model

TL;DR

This paper presents a method using a pre-trained autoregressive language model to extract cause-effect-signal triplets from news sentences, achieving competitive results despite limited training data.

Contribution

It introduces a novel approach for event causality extraction using iterative triplet prediction with a pre-trained language model, demonstrating effectiveness with minimal data.

Findings

Achieved second place in the CASE-2022 challenge.

Using cause-effect or effect-cause order yields similar performance.

Effective triplet extraction with only 160 training samples.

Abstract

In this paper, we describe our shared task submissions for Subtask 2 in CASE-2022, Event Causality Identification with Casual News Corpus. The challenge focused on the automatic detection of all cause-effect-signal spans present in the sentence from news-media. We detect cause-effect-signal spans in a sentence using T5 -- a pre-trained autoregressive language model. We iteratively identify all cause-effect-signal span triplets, always conditioning the prediction of the next triplet on the previously predicted ones. To predict the triplet itself, we consider different causal relationships such as cause$\rightarrow$effect$\rightarrow$signal. Each triplet component is generated via a language model conditioned on the sentence, the previous parts of the current triplet, and previously predicted triplets. Despite training on an extremely small dataset of 160 samples, our approach achieved competitive performance, being placed second in the competition. Furthermore, we show that assuming either cause$\rightarrow$effect or effect$\rightarrow$cause order achieves similar results.