Identifying emotions in social media: Comparison of word-emotion lexicons

In recent years, emotions expressed in social media messages have become a vivid research topic due to their influence on the spread of misinformation and online radicalization over online social networks. Thus, it is important to correctly identify emotions in order to make inferences from social media messages. In this paper, we report on the performance of three publicly available word-emotion lexicons (NRC, DepecheMood, EmoSenticNet) over a set of Facebook and Twitter messages. To this end, we designed and implemented an algorithm that applies natural language processing (NLP) techniques along with a number of heuristics that reflect the way humans naturally assess emotions in written texts. In order to evaluate the appropriateness of the obtained emotion scores, we conducted a questionnaire-based survey with human raters. Our results show that there are noticeable differences between the performance of the lexicons as well as with respect to emotion scores the human raters provided in our survey.