Clustering and sentiment analysis on Twitter data

Abstract

Twitter is a social media platform is a great place where people from all parts of the world can make their opinions heard. Twitter produces around 500 million of tweets daily which amounts to about 8TB of data. The data generated in twitter can be very useful if analyzed as we can extract important information via opinion mining. Opinions about any news or launch of a product or a certain kind of trend can be observed well in twitter data. The main aim of sentiment analysis (or opinion mining) is to discover emotion, opinion, subjectivity and attitude from a natural text. In twitter sentiment analysis, we categorize tweets into positive and negative sentiment. Clustering is a protean procedure in which identically resembled objects are grouped together and form a pack or cluster. We conducted a study and found out that the use of clustering can quickly and efficiently distinguish tweets on the basis of their sentiment scores and can find weekly and strongly positive or negative tweets when clustered with results of different dictionaries. This paper surveys different approaches of clustering with respect to sentiment analysis and presents a way to find relationships between the tweets on the basis of polarity and subjectivity.