Exploring emoticons in polarity classification of text

With people increasingly using emoticons in written text on the Web in order to express, stress, or disambiguate their sentiment, it is crucial for automated sentiment analysis tools to correctly account for such graphical cues for sentiment. We analyze how emoticons typically convey sentiment and we subsequently propose and evaluate a novel method for exploiting this with a manually created emoticon sentiment lexicon in a lexicon-based polarity classification method. We evaluate our approach on 2,080 Dutch tweets and forum messages, which all contain emoticons. We validate our findings on 10,069 English reviews of apps, some of which contain emoticons. We find that accounting for the sentiment conveyed by emoticons on a paragraph level -- and, to a lesser extent, on a sentence level -- significantly improves polarity classification performance. Whenever emoticons are used, their associated sentiment tends to dominate the sentiment conveyed by textual cues and forms a good proxy for the polarity of text.