Sentiment Analysis Based on Psychological and Linguistic Features for Spanish Language

Recent research activities in the areas of opinion mining, sentiment analysis and emotion detection from natural language texts are gaining ground under the umbrella of affective computing. Nowadays, there is a huge amount of text data available in the Social Media (e.g. forums, blogs, and social networks) concerning to users’ opinions about experiences buying products and hiring services. Sentiment analysis or opinion mining is the field of study that analyses people’s opinions and mood from written text available on the Web. In this paper, we present extensive experiments to evaluate the effectiveness of the psychological and linguistic features for sentiment classification. To this purpose, we have used four psycholinguistic dimensions obtained from LIWC, and one stylometric dimension obtained from WordSmith, for the subsequent training of the SVM, Naïve Bayes, and J48 algorithms. Also, we create a corpus of tourist reviews from the travel website TripAdvisor. The findings reveal that the stylometric dimension is quite feasible for sentiment classification. Finally, with regard to the classifiers, SVM provides better results than Naïve Bayes and J48 with an F-measure rate of 90.8%.