Emojis Sentiment Analysis Based on Big Social Media Data

As the use of emojis becomes ever-more prevalent in social media texts (e.g. twitters, microblogs), various kinds of work on emojis sentiment analysis have been carried out. However, there is little work calculating emojis sentiment quantificationally. This paper proposes an improved label propagation algorithm (LPA) called label attenuation propagation model (LAPM) to calculate emojis sentiment automatically, multi-dimensionally, and quantificationally. First, Emoji Link Network (ELN) is built to organize a huge number of emojis in social media. Second, LPA is introduced into the area of emojis sentiment analysis. Third, we measure the sentiment uncertainty of emojis and propose LAPM to calculate emojis sentiment. Experimental results show that the accuracy of LAPM is 85.2%, compared with 74.8% of LPA.