PSentScore: Evaluating Sentiment Polarity in Dialogue Summarization

TL;DR

This paper introduces PSentScore, a set of measures to evaluate how well dialogue summaries retain emotional content, revealing current models' shortcomings and how training data selection can improve affective preservation.

Contribution

The paper proposes PSentScore for assessing affective content preservation in dialogue summaries and shows how training data selection enhances this aspect.

Findings

State-of-the-art models poorly preserve affective content.

Careful training set selection improves affective content preservation.

Minor trade-off with content-related metrics.

Abstract

Automatic dialogue summarization is a well-established task with the goal of distilling the most crucial information from human conversations into concise textual summaries. However, most existing research has predominantly focused on summarizing factual information, neglecting the affective content, which can hold valuable insights for analyzing, monitoring, or facilitating human interactions. In this paper, we introduce and assess a set of measures PSentScore, aimed at quantifying the preservation of affective content in dialogue summaries. Our findings indicate that state-of-the-art summarization models do not preserve well the affective content within their summaries. Moreover, we demonstrate that a careful selection of the training set for dialogue samples can lead to improved preservation of affective content in the generated summaries, albeit with a minor reduction in content-related metrics.