Learning affective projections for emoticons on Twitter

Emoticons have in the literature been shown to modify rather than provide redundancy to the accompanying textual message. Despite this, emoticons are often used merely as labels for sentiment classification tasks. This paper aims to explore the phenomenon and discover more salient emoticon-emotion associations through an embedding-based machine learning process. Using principal component analysis and k-means clustering, it is shown how similar emoticons form groups in vector space. Furthermore, a supervised classification strategy for discovering emoticon-emotion associations is presented. A qualitative evaluation of the results shows that while the clustering is highly salient, the supervised approach does not perform as well.