Retrieving Implicit and Explicit Emotional Events Using Large Language Models

TL;DR

The work probes how well large language systems retrieve emotional events in commonsense contexts, addressing both implicit and explicit cues. It introduces a supervised contrastive probe to quantify precision and diversity of retrieved events across Joy, Sad, and Angry, using the C^3KG emotion-cause flow as a data source. Experiments reveal Joy is easiest to retrieve, while Sadness and Anger present greater challenges, and diversity remains limited without deduplication. The findings highlight strengths and limitations of current approaches for commonsense emotion reasoning and point to directions for achieving more balanced, robust emotion retrieval in future research.

Abstract

Large language models (LLMs) have garnered significant attention in recent years due to their impressive performance. While considerable research has evaluated these models from various perspectives, the extent to which LLMs can perform implicit and explicit emotion retrieval remains largely unexplored. To address this gap, this study investigates LLMs' emotion retrieval capabilities in commonsense. Through extensive experiments involving multiple models, we systematically evaluate the ability of LLMs on emotion retrieval. Specifically, we propose a supervised contrastive probing method to verify LLMs' performance for implicit and explicit emotion retrieval, as well as the diversity of the emotional events they retrieve. The results offer valuable insights into the strengths and limitations of LLMs in handling emotion retrieval.

Figures (1)

• Figure 1: Performance (P@3, P@10, P@50) of different LLMs on the emotion retrieval.