Assessment of feature selection metrics for sentiment analyses: Turkish movie reviews

Akba F., Uçan A., AKÇAPINAR SEZER E., Sever H.

European Conference on Data Mining 2014 and International Conferences on Intelligent Systems and Agents 2014 and Theory and Practice in Modern Computing 2014, Lisbon, Portugal, 15 - 17 July 2014, pp.180-184, (Full Text)

  • Publication Type: Conference Paper / Full Text
  • City: Lisbon
  • Country: Portugal
  • Page Numbers: pp.180-184
  • Keywords: Feature selection, Naïve bayes, Sentiment analyses, Support vector machine, Turkish corpus
  • TED University Affiliated: No

Abstract

Sentiment analysis systems pursuit the goal of detecting emotions in a given text with machine learning approaches. These texts might include three kinds of emotions such as positive, negative and neutral. Entertainment oriented texts, especially movie reviews, contain huge amount of possible emotional information. In this study, we aimed to represent each movie reviews by using small number of features. For this purpose, information gain, chi-square methods have been implemented to extract features for decreasing costs of calculations and increasing success rate. In experiments, employed corpus includes Turkish movie reviews, support vector machine and naïve bayes had been employed for classification and F1 score was used for performance evaluation. According to the experimental results, support vector machine achieved 83.9% performance value while classification of movie reviews in two (positive and negative) categories and also we obtained the 63.3% performance value while classification with support vector machine into three categories.