A Lexical Supervised Approach for Opinion Mining in the Domain of Laptops and Restaurants

Abstract

This paper presents a study of opinion mining or sentiment analysis for detection of polarity in a set of users opinion about restaurants written in Spanish and English. The research work is performed with the aim of solving a task proposed in SemEval 2016, thus we employed the same dataset proposed in that evaluation conference. The proposed approach uses a vector model for representing the information, including lexical features such as the following ones: word unigrams, bigrams and trigrams. The obtained results show a performance up to 71% when using word unigrams for representing the opinions written in English in the domain of restaurants.