This paper presents a method for using topic distributions generated from topic models as features for performing sentiment analysis on documents. This will be tested in the social media domain, specifically Twitter. The proposed approach allows for the mapping from word space to topic space which allows for less features to be needed and also reduces computational complexity. Multiple machine learning algorithms will be used to test the topic model generated features and a number of different versions of test corpus will be used, including unigrams, bigrams, part-of-speech tagging and adjectives only. The method proposed will also be compared to other notable topic-sentiment methods such as the aspect-sentiment unification model and the joint sentiment/topic model. The results show that using topic distributions can improve the accuracy of classification algorithms, however, the performance can be dependent on the algorithm used and the initial features used. Additionally, we show that using only topics as features outperforms the hybrid topic-sentiment models.
IOS Press, Inc.
6751 Tepper Drive
Clifton, VA 20124
Tel.: +1 703 830 6300
Fax: +1 703 830 2300 firstname.lastname@example.org
(Corporate matters and books only) IOS Press c/o Accucoms US, Inc.
For North America Sales and Customer Service
West Point Commons
Lansdale PA 19446
Tel.: +1 866 855 8967
Fax: +1 215 660 5042 email@example.com