Indian Institute of Technology-Patna: Sentiment Analysis in Twitter

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 341–345, Dublin, Ireland, August 23-24, 2014. Indian Institute of Technology-Patna: Sentiment Analysis in Twitter Vikram Singh, Arif Md. Khan and Asif Ekbal Indian Institute of Technology Patna Patna, India (vikram.mtcs13,arif.mtmc13,asif)@iitp.ac.in Abstract This paper is an overview of the system submitted to the SemEval-2014 shared task on sentiment analysis in twitter. For the very first time we participated in both the tasks, viz contextual polarity disam- biguation and message polarity classifi- cation. Our approach is supervised in nature and we use sequential minimal optimization classifier. We implement the features for sentiment analysis with- out using deep domain-specific resources and/or tools. Experiments within the benchmark setup of SemEval-14 shows the F-scores of 77.99%, 75.99%, 76.54 %, 76.43% and 71.43% for LiveJour- nal2014, SMS2013, Twitter2013, Twit- ter2014 and Twitter2014Sarcasm, respec- tively for Subtask A. For Subtask B we obtain the F-scores of 60.39%, 51.96%, 52.58%, 57.25%, 41.33% for five different test sets, respectively. 1 Introduction In current era microblogging is an efficient way of communication where people can communicate without physical presence of receiver(s). Twitter is the medium where people post real time mes- sages to discuss on the different topics, and ex- press their sentiments. The texts used in twit- ter are generally informal and unstructured in na- ture. Tweets and SMS messages are very short in length, usually a sentence or a headline rather than a document. These texts are very informal in nature and contains creative spellings and punc- tuation symbols. Text also contains lots of mis- spellings, slang, out-of-vocabulary words, URLs, 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/ and genre-specific terminology and abbreviations, e.g., RT for re-Tweet and #hashtags. Such kinds of structures introduce difficulties in building various lexical and syntactic resources and/or tools, which are required for efficient processing of texts. Find- ing relevant information from these posts poses big challenges to the researchers compared to the traditional text genres such as newswire. In recent times, there has been a huge inter- est to mine and understand the opinions and sen- timents that people are communicating in social media (Barbosa and Feng, 2010; Bifet et al., ; Pak and Paroubek, 2010; Kouloumpis et al., 2011). There is a tremendous interest in sentiment analysis of Tweets across a variety of domains such as commerce (Jansen et al., 2009), health (Chew and Eysenbach, 2010; Salathe and Khan- delwal, 2011) and disaster management (Verma et al., 2011; Mandel et al., 2012). Agarwal et al.(Agarwal et al., 2011) used tree kernel decision tree that made use of the features such as Part- of-Speech (PoS) information, lexicon-based fea- tures and several other features. They acquired 11,875 manually annotated Twitter data (Tweets) from a commercial source, and reported an accu- racy of 75.39%. Semantics has also been used as the feature to improve the performance of senti- ment analysis (Saif et al., …

Content truncated to 500 words