Minería de opinión para textos en español usando procesamiento natural del lenguaje

Abstract

This paper proposes development of a tool that allows opinion mining of texts exclusively in Spanish. For this, a linguistic corpus has been constructed through tweets. This corpus is characterized by having tweets in Latin American Spanish and Castilian Spanish. The evaluated classification algorithms were: Naive Bayes, SVM, and LSTM. The experiments were carried out with a corpus of 14,666 with a classification in two classes (positive and negative) in three classes (positive, negative, neutral). The results obtained show that the binary classification got better results than the three-class classification in the three algorithms.