Author(s)

KAREL GUTIERREZ-BATISTA, JESÚS R. CAMPAÑA, SANDRO MARTINEZ-FOLGOSO, M. AMPARO VILA & MARIA J. MARTIN-BAUTISTA

Abstract

Topic detection from large textual data volumes extracted from Social Networks is an interesting research topic in the context of Big Data. The textual content present in Social Networks contains diverse information that can be exploited in order to obtain useful information. Topic detection and sentiment analysis in social networks are topics of widespread research. The study of both is sometimes intertwined as, usually, user messages revolve around a particular topic and express certain attitude of the user towards the topic discussed. However, this assumption is not valid for all messages as some of them express only general feelings or attitudes and do not refer to something in particular that covers up the topic discussed. In fact, these messages can influence the topic detection process. In this paper, we propose to obtain topics from massive quantities of text data extracted from social networks, without using previous information, and only with the use of unsupervised data mining techniques. We analyze the influence of sentiments in messages and how they affect the topic detection task. Terms related to sentiments provide useful information for a variety of applications, but not for topic detection where they represent a source of unnecessary noise. Experiments are conducted on data obtained from Twitter social network

Keywords