Automatic identification of sentiment in unstructured text

Abstract

The constant increase of information in digital format forces us to have new tools that allow us to download, organize and analyze the information available on the web. One of the analyses performed on unstructured information is polarity identification. In this paper we present a method to carry out polarity identification in unstructured texts. Specifically, texts downloaded from the social network Twitter are used. The current popularity of social networks, has caused a great prominence among different users for the generation of information day by day. Twitter presents us with a great challenge in the automatic processing of natural language, mainly when the number of opinions is very large and automatic processing is required. In our case, in the determination of the polarity contained in a tweet. In this paper we present results obtained using different machine learning methods widely known in the state of the art, such as: Support Vector Machine, Naive Bayes, Logistic Regression, Nearest Neighbors and Random Forest, which are used in two implemented classification scenarios: cross-validation and training and test sets. Two data sets are used for the evaluation of the implemented methodology. The best results are obtained with Support Vector Machine for both datasets, the obtained accuracy values higher than 83 % allow to see the viability of the implemented methodology.