Abstract:
Classifying polarity of sentiments expressed in micro-blogs, such as tweets, is an active research area nowadays. The research direction has been focusing on classifying sentiments towards specific targets, i.e., topics, in the micro-blog. A more recent direction currently addresses the problem of detecting the target then identifying the sentiment toward it. While the former direction is referred to as target-dependent sentiment classification, the latter direction is referred to as open domain targeted sentiment classification. Many approaches have been proposed in the literature for automatic sentiment classification. Most of these approaches use supervised learning techniques that exploit only labeled data for training their proposed models. This paper presents an invited extension to a recent survey published by the authors. In this paper, we compile and present the accuracy reported by researchers with respect to the application of different techniques when applied to the same dataset. Our study presents comparisons between different techniques with regard to both the target-dependent and the open domain targeted sentiment classification. The study identifies some gaps to be addressed in future research. For instance, it shows that performance of both target-dependent and open domain targeted sentiment classification is still limited, and further future research could be promising.
