MHSubLex: Using Metaheuristic Methods for Subjectivity Classification of Microblogs

In Web 2.0, people are free to share their views, experiences, and opinions. One of the problems that arise in Web 2.0 is the sentiment analysis of texts produced by users in outlets such as Twitter. One of the main tasks of sentiment analysis is subjectivity classification. Our aim is to classify the subjectivity of tweets. To this end, we create subjectivity lexicons, in which the words are classified into the objective and subjective ones. To create these lexicons, we make use of three metaheuristic methods. Two meta-level features are extracted in this method, which show the number of subjective and objective words in tweets according to the lexicons. We then classify the records based upon these two features. By comparing accuracy and f-measure to the baselines, it is shown that the proposed method performs better. In the three metaheuristics, it is observed that the genetic algorithm performs better than the simulated annealing and the asexual reproduction optimization, and its performance is also better than all baselines in two of the three assessed datasets, in terms of accuracy. The lexicons that are created using this method can give an insight about the subjectivity and objectivity of words.