Fine-grained Sentiment Analysis with Faithful Attention

TL;DR

This paper improves fine-grained sentiment analysis between specific source and target entities by training model attention with human rationales, leading to better performance and more faithful explanations.

Contribution

It introduces a method to train attention with human rationales, significantly enhancing both model accuracy and interpretability in sentiment analysis.

Findings

Training attention with human rationales improves performance by 4-8 points.

Untrained attention is often misaligned and not faithful.

A small amount of human rationales suffices to correct attention.

Abstract

While the general task of textual sentiment classification has been widely studied, much less research looks specifically at sentiment between a specified source and target. To tackle this problem, we experimented with a state-of-the-art relation extraction model. Surprisingly, we found that despite reasonable performance, the model's attention was often systematically misaligned with the words that contribute to sentiment. Thus, we directly trained the model's attention with human rationales and improved our model performance by a robust 4~8 points on all tasks we defined on our data sets. We also present a rigorous analysis of the model's attention, both trained and untrained, using novel and intuitive metrics. Our results show that untrained attention does not provide faithful explanations; however, trained attention with concisely annotated human rationales not only increases performance, but also brings faithful explanations. Encouragingly, a small amount of annotated human rationales suffice to correct the attention in our task.