Evaluation of Four Approaches for "Sentiment Analysis on Movie Reviews"
Athanasia KoumpouriDept. of Computer Engineering and Informatics, University of Patras, 26500-Rion, Greece, 0030 2610997534Iosif MporasDept. of Computer Engineering and Informatics, University of Patras, 26500-Rion, Greece, 0030 2610997534Vasileios MegalooikonomouDept. of Computer Engineering and Informatics, University of Patras, 26500-Rion, Greece, 0030 2610997534
2015en
ABI
Аннотация
In this paper we present four different approaches for automatic sentiment classification on movie reviews. The proposed approaches: (a) statistical based, (b) bag-of-words based, (c) content based and (d) lexicon based approach, were evaluated in the "Sentiment Analysis on Movie Reviews" Kaggle competition. The competition results showed that each of the first three aforementioned approaches achieved approximately 65% sentiment classification accuracy, while the latter performed poorly compared to the others. The combination scheme of (b)-(c)-(d) proved to be our best performing set up achieving our best classification accuracy 67.931%, which ranked us in the 6th position among 861 participants.
Перевод пока недоступен
Идентификаторы
Цитирования и источники
Цитирований: 2Использованных источников: 0