CMUQ@Qatar:Using Rich Lexical Features for Sentiment Analysis on Twitter

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 186–191, Dublin, Ireland, August 23-24, 2014. CMUQ@Qatar:Using Rich Lexical Features for Sentiment Analysis on Twitter Sabih Bin Wasi, Rukhsar Neyaz, Houda Bouamor, Behrang Mohit Carnegie Mellon University in Qatar {sabih, rukhsar, hbouamor, behrang}@cmu.edu Abstract In this paper, we describe our system for the Sentiment Analysis of Twitter shared task in SemEval 2014. Our system uses an SVM classifier along with rich set of lexical features to detect the sentiment of a phrase within a tweet (Task-A) and also the sentiment of the whole tweet (Task- B). We start from the lexical features that were used in the 2013 shared tasks, we en- hance the underlying lexicon and also in- troduce new features. We focus our fea- ture engineering effort mainly on Task- A. Moreover, we adapt our initial frame- work and introduce new features for Task- B. Our system reaches weighted score of 87.11% in Task-A and 64.52% in Task-B. This places us in the 4th rank in the Task- A and 15th in the Task-B. 1 Introduction With more than 500 million tweets sent per day, containing opinions and messages, Twitter1 has become a gold-mine for organizations to monitor their brand reputation. As more and more users post about products and services they use, Twit- ter becomes a valuable source of people’s opin- ions and sentiments: what people can think about a product or a service, how positive they can be about it or what would people prefer the product to be like. Such data can be efficiently used for mar- keting. However, with the increasing amount of tweets posted on a daily basis, it is challenging and expensive to manually analyze them and locate the meaningful ones. There has been a body of re- cent work to automatically learn the public sen- This work is licensed under a Creative Commons At- tribution 4.0 International Licence. Page numbers and pro- ceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ 1http://twitter.com timents from tweets using natural language pro- cessing techniques (Pang and Lee, 2008; Jansen et al., 2009; Pak and Paroubek, 2010; Tang et al., 2014). However, the task of sentiment analysis of tweets in their free format is harder than that of any well-structured document. Tweet messages usu- ally contain different kinds of orthographic errors such as the use of special and decorative charac- ters, letter or word duplication, extra punctuation, as well as the use of special abbreviations. In this paper, we present our machine learn- ing based system for sentiment analysis of Twitter shared task in SemEval 2014. Our system takes as input an arbitrary tweet and assigns it to one of the following classes that best reflects its sen- timent: positive, negative or neutral. While pos- itive and negative tweets are subjective, neutral class encompasses not only objective tweets but also subjective tweets that does not contain any ”polar” emotion. Our classifier was developed as an undergrad course project but later pursued as …

Content truncated to 500 words