Sentiment-Aware Extractive and Abstractive Summarization for Unstructured Text Mining
Junyi Liu, Stanley Kok
TL;DR
This work tackles the challenge of summarizing emotion-rich, unstructured text in Information Systems by proposing sentiment-aware extensions to extractive and abstractive paradigms. It introduces ECPE-TextRank and Senti-UniLM, which integrate emotion signals via emotion-cause extraction, attribution-based salience, and sentiment-weighted generation. Empirical results on Reddit-TIFU and DialogSum show clear improvements over strong baselines, with ablation analyses validating the contributions of sentiment and topic modules. The approach offers practical benefits for rapid brand monitoring and market analysis in dynamic online environments and outlines directions for domain adaptation and finer-grained emotion modeling.
Abstract
With the rapid growth of unstructured data from social media, reviews, and forums, text mining has become essential in Information Systems (IS) for extracting actionable insights. Summarization can condense fragmented, emotion-rich posts, but existing methods-optimized for structured news-struggle with noisy, informal content. Emotional cues are critical for IS tasks such as brand monitoring and market analysis, yet few studies integrate sentiment modeling into summarization of short user-generated texts. We propose a sentiment-aware framework extending extractive (TextRank) and abstractive (UniLM) approaches by embedding sentiment signals into ranking and generation processes. This dual design improves the capture of emotional nuances and thematic relevance, producing concise, sentiment-enriched summaries that enhance timely interventions and strategic decision-making in dynamic online environments.