Beneath the Tip of the Iceberg: Current Challenges and New Directions in Sentiment Analysis Research

Sentiment analysis as a field has come a long way since it was first\nintroduced as a task nearly 20 years ago. It has widespread commercial\napplications in various domains like marketing, risk management, market\nresearch, and politics, to name a few. Given its saturation in specific\nsubtasks -- such as sentiment polarity classification -- and datasets, there is\nan underlying perception that this field has reached its maturity. In this\narticle, we discuss this perception by pointing out the shortcomings and\nunder-explored, yet key aspects of this field that are necessary to attain true\nsentiment understanding. We analyze the significant leaps responsible for its\ncurrent relevance. Further, we attempt to chart a possible course for this\nfield that covers many overlooked and unanswered questions.\n