Sentiment-based influence detection on Twitter

Abstract

Abstract The user generated content available in online communities is easy to create and consume. Lately, it also became strategically important to companies interested in obtaining population feedback on products, merchandising, etc. One of the most important online communities is Twitter: recent statistics report 65 million new tweets each day. However, processing this amount of data is very costly and a big portion of the content is simply not useful for strategic analysis. Thus, in order to filter the data to be analyzed, we propose a new method for ranking the most influential users in Twitter. Our approach is based on a combination of the user position in networks that emerge from Twitter relations, the polarity of her opinions and the textual quality of her tweets. Our experimental evaluation shows that our approach can successfully identify some of the most influential users and that interactions between users provide the best evidence to determine user influence.