DIEGOLab: An Approach for Message-level Sentiment Classification in Twitter

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 510–514, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics DIEGOLab: An Approach for Message-level Sentiment Classification in Twitter Abeed Sarker, Azadeh Nikfarjam, Davy Weissenbacher, Graciela Gonzalez Department of Biomedical Informatics Arizona State University Scottsdale, AZ 85281, USA {abeed.sarker,anikfarj,dweissen,graciela.gonzalez}@asu.edu Abstract We present our supervised sentiment classifi- cation system which competed in SemEval- 2015 Task 10B: Sentiment Classification in Twitter— Message Polarity Classification. Our system employs a Support Vector Ma- chine classifier trained using a number of fea- tures including n-grams, dependency parses, synset expansions, word prior polarities, and embedding clusters. Using weighted Sup- port Vector Machines, to address the issue of class imbalance, our system obtains positive class F-scores of 0.701 and 0.656, and nega- tive class F-scores of 0.515 and 0.478 over the training and test sets, respectively. 1 Introduction Social media has seen unprecedented growth in recent years. Twitter, for example, has over 645,750,000 users and grows by an estimated 135,000 users every day, generating 9,100 tweets per second1). Users often express their views and emotions regarding a range of topics on social me- dia platforms. As such, social media has become a crucial resource for obtaining information directly from end-users, and data from social media has been utilized for a variety of tasks ranging from person- alized marketing to public health monitoring. While the benefits of using a resource such as Twitter in- clude large volumes of data and direct access to end- user sentiments, there are several obstacles associ- ated with the use of social media data. These include 1http://www.statisticbrain.com/ twitter-statistics/. Accessed on: 26th August, 2014. the use of non-standard terminologies, misspellings, short and ambiguous posts, and data imbalance, to name a few. In this paper, we present a supervised learning approach, using Support Vector Machines (SVMs) for the task of automatic sentiment classification of Twitter posts. Our system participated in the SemEval-2015 task Sentiment Classification in Twitter— Message Polarity Classification. The goal of the task was to automatically classify the polar- ity of a Twitter post into one of three predefined categories— positive, negative and neutral. In our approach, we apply a small set of carefully extracted lexical, semantic, and distributional features. The features are used to train a SVM learner, and the issue of data imbalance is addressed by using dis- tinct weights for each of the three classes. The re- sults of our system are promising, with positive class F-scores of 0.701 and 0.656, and negative class F- scores of 0.515 and 0.478 over the training and test sets, respectively. 2 Related Work Following the pioneering work on sentiment analy- sis by Pang et. al. (2002), similar research has been carried out under various umbrella terms such as: se- mantic orientation (Turney, 2002), opinion mining (Pang and Lee, 2008), polarity classification (Sarker et al., 2013), and many more. Pang et al. (2002) uti- lized machine learning models to predict sentiments in text, and their approach showed that SVM clas- sifiers …

Content truncated to 500 words