An Analysis of Influence of Emoticons on Affective Impressions Feeling from Tweets

In this paper, we investigated how Twitter users perceived affective impressions from tweets (with and without emoticons), and formulated the influence of emoticons on affective impressions. Initially, we conducted questionnaires, and quantified the impressions associated with three types of text: tweets with emoticons, tweets without emoticons, and emoticons. Multiple regression analysis was then applied to the three types of impression data, and consequently, multiple regression equations representing the relationships among them were obtained, where impression data on the tweets with emoticons were used as the objective variable, and impression data on the tweets without emoticons and the emoticons were used as the explanatory variables. Finally, the accuracy of the equations was estimated for learned and unlearned data, and their effectiveness was shown. Note that our target impressions are limited to the following eight types: "Offensive and/or Unpleasant," "Negative," "Good feeling," "Happy and/or Pleasant," "Positive," "Warm feel," "Gloomy," and "Scary."