The Impact of Z_score on Twitter Sentiment Analysis

Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), pages 636–641, Dublin, Ireland, August 23-24, 2014. The Impact of Z_score on Twitter Sentiment Analysis Hussam Hamdan*,**,*** *LSIS Aix-Marseille Université CNRS Av. Esc. Normandie Niemen, 13397 Marseille Cedex 20, France hussam.hamdan@lsis.org Patrice Bellot*,** **OpenEdition Aix-Marseille Université CNRS 3 pl. V. Hugo, case n°86 13331 Marseille Cedex 3, France patrice.bellot@lsis.org Frederic Béchet*** ***LIF Aix-Marseille Université CNRS Avenue de Luminy 13288 Marseille Cedex 9, France frederic.bechet@lif.univ-mrs.fr Abstract Twitter has become more and more an im- portant resource of user-generated data. Sen- timent Analysis in Twitter is interesting for many applications and objectives. In this pa- per, we propose to exploit some features which can be useful for this task; the main contribution is the use of Z-scores as features for sentiment classification in addition to pre-polarity and POS tags features. Our ex- periments have been evaluated using the test data provided by SemEval 2013 and 2014. The evaluation demonstrates that Z_scores features can significantly improve the predic- tion performance. 1 Introduction The interactive Web has changed the relation between the users and the web. Users have be- come an important source of content. They ex- press their opinion towards different issues. The- se opinions are important for others who are in- terested in understanding users’ interests such as buyers, sellers and producers. Twitter is one of the most important platforms in which the users express their opinions. Many works have exploited this media for predicting valuable issues depending on Sentiment Analysis (SA). The authors in (Asur and Huberman 2010) predicted the box-office revenues of movies in advance of their releases using the tweets talking about them. In (Bae and Lee 2012) Sentiment This work is licensed under a Creative Commons At- tribution 4.0 International Licence. Page numbers and proceedings footer are added by the organisers. Licence details: http://creativecommons.org/licenses/by/4.0/ Analysis has been used to study the impact of 13 twitter accounts of famous persons on their fol- lowers and also for forecasting the interesting tweets which are more probably to be reposted by the followers (Naveed, Gottron et al. 2011). Sentiment Analysis can be done in different lev- els; Document level; Sentence level; Clause level or Aspect-Based level. SA in Twitter can be seen as a sentence level task, but some limitations should be considered in such sentences. The size of tweets is limited to 140 characters, informal language, emotion icons and non-standard ex- pressions are commonly used, and many spelling errors can be found due to the absence of cor- rectness verification. Three different approaches can be identified in the literature of Sentiment Analysis in Twitter, the first approach is lexicon based, using specific types of lexicons to derive the polarity of a text, this approach suffers from the limited size of lex- icon and requires human expertise to build man- ual lexicon (Joshi, Balamurali et al. 2011), in the other hand the automatic lexicons are not so effi- cient. The second one is machine learning ap- proach which uses annotated …

Content truncated to 500 words