Minimally Supervised Learning of Affective Events Using Discourse Relations

TL;DR

This paper introduces a simple discourse relation-based method for affective event recognition that effectively propagates sentiment polarity using minimal supervision, improving performance especially with limited labeled data.

Contribution

It presents a novel minimally supervised approach leveraging discourse relations to identify affective events without extensive labeled data.

Findings

Effective propagation of affective polarity using discourse relations

Improved affective event recognition with small labeled datasets

Successful application to Japanese data

Abstract

Recognizing affective events that trigger positive or negative sentiment has a wide range of natural language processing applications but remains a challenging problem mainly because the polarity of an event is not necessarily predictable from its constituent words. In this paper, we propose to propagate affective polarity using discourse relations. Our method is simple and only requires a very small seed lexicon and a large raw corpus. Our experiments using Japanese data show that our method learns affective events effectively without manually labeled data. It also improves supervised learning results when labeled data are small.