Polarity and Intensity: the Two Aspects of Sentiment Analysis
This work addresses sentiment analysis for researchers by providing a more nuanced understanding of sentiment scores, though it is incremental as it builds on existing multimodal frameworks.
The authors tackled the problem of sentiment analysis by decomposing sentiment scores into polarity and intensity aspects, and found that multi-task learning models improved sentiment analysis performance while revealing differences in how individual modalities convey these aspects.
Current multimodal sentiment analysis frames sentiment score prediction as a general Machine Learning task. However, what the sentiment score actually represents has often been overlooked. As a measurement of opinions and affective states, a sentiment score generally consists of two aspects: polarity and intensity. We decompose sentiment scores into these two aspects and study how they are conveyed through individual modalities and combined multimodal models in a naturalistic monologue setting. In particular, we build unimodal and multimodal multi-task learning models with sentiment score prediction as the main task and polarity and/or intensity classification as the auxiliary tasks. Our experiments show that sentiment analysis benefits from multi-task learning, and individual modalities differ when conveying the polarity and intensity aspects of sentiment.