Zero-shot Temporal Relation Extraction with ChatGPT

TL;DR

This paper evaluates ChatGPT's zero-shot ability to extract temporal relations between events in documents, highlighting its strengths in small relation classes and current limitations in consistency and long-dependency inference.

Contribution

It introduces prompt techniques for zero-shot temporal relation extraction and compares ChatGPT's performance with supervised methods.

Findings

ChatGPT performs better on small relation classes than supervised models.

Prompt design significantly affects ChatGPT's extraction accuracy.

ChatGPT struggles with maintaining consistency and long-dependency inference.

Abstract

The goal of temporal relation extraction is to infer the temporal relation between two events in the document. Supervised models are dominant in this task. In this work, we investigate ChatGPT's ability on zero-shot temporal relation extraction. We designed three different prompt techniques to break down the task and evaluate ChatGPT. Our experiments show that ChatGPT's performance has a large gap with that of supervised methods and can heavily rely on the design of prompts. We further demonstrate that ChatGPT can infer more small relation classes correctly than supervised methods. The current shortcomings of ChatGPT on temporal relation extraction are also discussed in this paper. We found that ChatGPT cannot keep consistency during temporal inference and it fails in actively long-dependency temporal inference.