EmotionExpert: Facebook game for crowdsourcing annotations for emotion detection

Abstract

The current paper explores the use of the social network platform Facebook, as a source of emotion annotated textual data as well as a source of annotators. The traditional approach of hiring experts to provide manually labeled (annotated) data for NLP research is time-consuming, tedious and expensive. Hence, crowdsourcing has emerged as a useful method for obtaining annotated data for natural language processing (NLP) research. We have developed a purposeful innovative Facebook game called EmotionExpert in order to collect human annotated textual data for emotion detection from text. The game provides a means to reach a large number of players, while making the annotation of emotional content of texts an enjoyable and social activity. The findings reported in this paper indicate that EmotionExpert is a useful resource for reaching a large number of people to produce reliable annotations.