TJP: Identifying the Polarity of Tweets from Contexts

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 657–662, Dublin, Ireland, August 23-24, 2014. TJP: Identifying the Polarity of Tweets from Context Tawunrat Chalothorn Jeremy Ellman Department of Computer Science and Digital Technologies University of Northumbria at Newcas- tle, Pandon Building, Camden Street Newcastle Upon Tyne, NE2 1XE, UK Department of Computer Science and Digital Technologies University of Northumbria at Newcas- tle, Pandon Building, Camden Street Newcastle Upon Tyne, NE2 1XE, UK tawunrat.chalothorn @northumbria.ac.uk jeremy.ellman @northumbria.ac.uk Abstract The TJP system is presented, which partici- pated in SemEval 2014 Task 9, Part A: Contextual Polarity Disambiguation. Our system is ‘constrained’, using only data provided by the organizers. The goal of this task is to identify whether marking contexts are positive, negative or neutral. Our system uses a support vector machine, with exten- sive pre-processing and achieved an overall F-score of 81.96%. 1 Introduction The aim of sentiment analysis is to identify whether the subject of a text is intended to be viewed positively of negatively by a reader. Such emotions are sometimes hidden in long sentences and are difficult to identify. Consequently senti- ment analysis is an active research area in natural language processing. * Sentiment is currently conceived terms of po- larity. This has numerous interesting applica- tions. For example, Grabner et al. (2012) used sentiment analysis to classify customers’ reviews of hotels by using a star rating to categorize the * 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/ reviews as bad, neutral and good. Similarly, Tumasjan et al. (2010) tried to predict the out- come of the German federal election through the analysis more than 100,000 tweets posted in the lead up. Sentiment analysis has also used to classify whether dreams are positive or nega- tive (Nadeau et al. 2006). This paper presents the TJP system which was submitted to SemEval 2014 Task 9, Part A: Contextual Polarity Disambiguation (Rosenthal et al., 2014). TJP focused on the ‘Constrained’ task. The ‘Constrained’ task only uses data provid- ed by the organizers. That is, external resources such as sentiment inventories (e.g. Sentiwordnet (Esuli, and Sebastiani 2006) are excluded. The objective of the TJP system was to use the results for comparison with our previous experiment (Chalothorn and Ellman, 2013). More details of these can be found in section 5. The TJP system was implemented using a support vector machine (SVM, e.g. Joachims, 1999) with the addition of extensive pre- processing such as stopword removal, negation, slang, contraction, and emoticon expansions. The remainder of this paper is constructed as follows: firstly, related work is discussed in sec- tion 2; the methodology, the experiment and re- sults are presented in sections 3 and 4, 657 respectively. Finally a discussion and future work are given in section 5. 2 Related Work Twitter is a popular social networking and mi- croblogging site that allows users to post mes- sages of up …

Content truncated to 500 words