Author:

Abstract:

Due to the increasing popularity of social networks, emotional data is generated more than ever. Feelings, opinions, experiences and comments are expressed, whether or not explicitly, toward products, services, brands, events etc. Understanding or analysing these sentiments in an automatic way can leverage valuable knowledge for businesses, governments or individuals. Sentiment analysis is therefore becoming an exciting subfield of data mining, coined as emotion mining. In this paper we present an architecture to automatically extract the emotional connotation of twitters' microblogs in the context of events. This entails finding and filtering tweets based on their relevance, attaching emotional scores to words inside tweets and combining these scores into some global emotional evaluation. As an ontology, we will examine the strength of SENTIWORDNET to attach emotional scores to the words inside tweets. Based on these scores, we propose the notion of an emotional distance measurement. This measurement makes it possible to calculate distances between tweets and cluster them based on their emotional value. Our architecture is tested on the Betagroup conference series.