Combining sentiment analysis and social network analysis to explore twitter opinion spreading

In this work, we reconstruct the tweet-retweet and tweet-reply relations of opinions about a trending topic on the Twitter platform. We propose a multi-steps approach to derive a signed network expressing the spread of contents and opinions. The first step consists in reducing data dimensionality by means of a clustering procedure on tweets able to identify the concepts they express. In the second step, focusing on message contents, we adapt different sentiment analysis algorithms in order to determine the sign of both the original tweet (with respect to the trending topic) and the sign of the edge connecting the original tweet to the replies, conditional on the replied tweet. Each tweet will spread its concepts by means of signed retweet and reply relations. The aim is to study the different structure, in terms of both network structure and sentiment, of the signed network related to each concept. A comparative analysis will be possible as well among the various identified signed networks.