顔文字がツイート感情の推測に及ぼす影響の定式化

-ツイート閲覧者が推測するツイート投稿者の感情を対象として-

抄録

In this paper, we investigated how readers perceive the emotions of writers based on the emoticons in their tweets, and clarified the influence of emoticons on emotions. Initially, we conducted a questionnaire and the emotions associated with tweets (with and without emoticons) and the emotions associated with the emoticons were quantified. Multiple regression analysis was then applied to three types of emotion data, and relationships among them were established in the form of regression equations. Finally, the accuracy of the regression equations was estimated for learned and unlearned data, and their effectiveness and robustness were evaluated. The following 10 types of basic emotions were utilized as targets: “Sadness,” “Dislike,” “Relief,” “Fear,” “Excitement,” “Liking,” “Joy,” “Surprise,” “Anger,” and “Shame.”