sielers : Feature Analysis and Polarity Classification of Expressions from Twitter and SMS Data

Second Joint Conference on Lexical and Computational Semantics (*SEM), Volume 2: Seventh International Workshop on Semantic Evaluation (SemEval 2013), pages 525–529, Atlanta, Georgia, June 14-15, 2013. c⃝2013 Association for Computational Linguistics sielers : Feature Analysis and Polarity Classification of Expressions from Twitter and SMS Data Harshit Jain, Aditya Mogadala and Vasudeva Varma Search and Information Extraction Lab, IIIT-H Hyderabad India harshit.jain@research.iiit.ac.in, aditya.m@research.iiit.ac.in, vv@iiit.ac.in Abstract In this paper, we describe our system for the SemEval-2013 Task 2, Sentiment Analysis in Twitter. We formed features that take into ac- count the context of the expression and take a supervised approach towards subjectivity and polarity classification. Experiments were per- formed on the features to find out whether they were more suited for subjectivity or po- larity Classification. We tested our model for sentiment polarity classification on Twitter as well as SMS chat expressions, analyzed their F-measure scores and drew some interesting conclusions from them. 1 Introduction In recent years there has been a huge growth in pop- ularity of vaious social media microblogging plat- forms like Twitter. Users freely share their personal opinions on various events and entities on these plat- forms. However, while character constraints make sure the opinions are short and to the point, they also contribute to the noisy nature of Twitter data. The contextual polarity of the phrase in which a particular instance of a word appears may be quite different from the word’s prior polarity. Positive words are used in phrases expressing negative sen- timents, or vice versa. Also, quite often words that are positive or negative out of context are neutral in context, meaning they are not even being used to ex- press a sentiment. This is evident from the example of underlined phrase in the following tweet: Lana Del Rey at Hammersmith Apollo in May...Very badly want tickets In a technique with large lexicon of words marked with their prior polarity, badly would have a negative score making the whole sentence with negative sen- timent. Even if we perform phrase-level analysis for the phrase “Very badly”, Very only acts as an intensi- fier for badly and the whole sentence is still marked negative. It’s only when we look further from the underlined phrase that we realize that “Very badly” in the context of wanting something shows positive sentiment. Early work on sentiment analysis is based on document-level analysis of reviews (Pang, B., and Lee, L., 2004). This approach isn’t feasible for mi- croblogging data due to the extremely small size of individual documents. The results on the effective- ness of part-of-speech features are mixed. While most regard POS features helpful in subjectivity classification (Barbosa, L. and Feng, J., 2010), some report very insignificant improvement on using them (Kouloumpis, E., Wilson, T. and Moore, J., 2011). However, most phrase-level approaches began with a large lexicon of words marked with their prior po- larity (Kim, S. M., and Hovy, E., 2004; Hu, M., and Liu, B, 2004). Wilson, Wiebe and Hoffman (2005) sought to include contextual polarity in …

Content truncated to 500 words