A Lexicon Approach to Multidimensional Analysis of Tweets Opinion

Nowadays, social media present a valuable source for business decision support. This article outlines the integration of social opinion data in multidimensional design combining sentiment analysis techniques and ETL design to offer a novel approach for social ETL design. The main contribution of this paper is the definition of a lexicon opinion analysis approach that extracts the sentiment polarity of informal text expressed in the Twitter social network. We propose a new algorithm, POLSentiment, based on lexical resources to firstly extract opinion words and emoticons from the tweet and then detect its positive or negative polarity. To assess the performance of the proposed algorithm, we evaluate POLSentiment on Sanders dataset. The results show that the proposal is suitable to automate the whole polarity analysis process, providing high accuracy levels and low false positive rates. Additionally, we define ETL processes design that consists of extracting tweets, its preprocessing, opinion analysis and polarity classification, then its loading into the Social Data Webhouse.