GTI at SemEval-2016 Task 4: Training a Naive Bayes Classifier using Features of an Unsupervised System

Proceedings of SemEval-2016, pages 115–119, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics GTI at SemEval-2016 Task 4: Training a Naive Bayes Classifier using Features of an Unsupervised System Jonathan Juncal-Mart´ınez, Tamara ´Alvarez-L´opez, Milagros Fern´andez-Gavilanes Enrique Costa-Montenegro, Francisco Javier Gonz´alez-Casta˜no GTI Research Group AtlantTIC Centre, School of Telecommunication Engineering, University of Vigo 36310 Vigo, Spain {jonijm, talvarez, milagros.fernandez, kike}@gti.uvigo.es, javier@det.uvigo.es Abstract This paper presents the approach of the GTI Research Group to SemEval-2016 task 4 on Sentiment Analysis in Twitter, or more specif- ically, subtasks A (Message Polarity Classifi- cation), B (Tweet classification according to a two-point scale) and D (Tweet quantification according to a two-point scale). We followed a supervised approach based on the extraction of features by a dependency parsing-based ap- proach using a sentiment lexicon and Natural Language Processing techniques. 1 Introduction In recent years, research on the field of Sentiment Analysis (SA) has increased considerably, due to the growth of user content generated in social networks, blogs and other platforms on the Internet. These are considered valuable information for companies, which seek to know or even predict the acceptance of their products, to design their marketing cam- paigns more efficiently. One of these sources of in- formation is Twitter, where users can write about any topic, using colloquial and compact language. As a consecuence, SA in Twitter is specially chal- lenging, as opinions are expressed in one or two short sentences. Many approaches have been proposed for SA, and can be roughly divided into two categories. The first one tries to capture and model linguistic knowl- edge through the use of dictionaries (Taboada et al., 2011) containing words that are tagged with their se- mantic orientation. These methods detect the words present in a text using different strategies involving lexics, syntax or semantics (Quinn et al., 2010). The other one is machine learning-based, which is cur- rently the most predominant approach including su- pervised learning and deep learning. They widely use classifiers including Support Vector Machines (SVM), Maximum Entropy Models (MAXENT), and Naive Bayes classifiers. Most of the time, they are built from features of a “bag of words” representa- tion (Pak and Paroubek, 2010). Our group has participated in SemEval-2016 task 4 on Sentiment Analysis in Twitter, sub- tasks A (Message Polarity Classification), B (Tweet classification according to a two-point scale) and D (Tweet quantification according to a two-point scale) (Nakov et al., 2016b). The remainder of this article is structured as fol- lows: Section 2 presents in detail the system pro- posed for the performance of these subtasks, and Section 3 shows the results obtained and discusses them. Finally, Section 4 summarizes the main find- ings and conclusions. 2 System Overview Our main objective was to create a supervised sys- tem using extracted features from an unsupervised system described in (Fern´andez-Gavilanes et al., 2015). This last approach comprises different pro- cessing stages, including the generation of senti- ment lexicons, test preprocessing and the applica- tion of different methods for determining contex- …

Content truncated to 500 words