LyS at SemEval-2016 Task 4: Exploiting Neural Activation Values for Twitter Sentiment Classification and Quantification

Proceedings of SemEval-2016, pages 79–84, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics LyS at SemEval-2016 Task 4: Exploiting Neural Activation Values for Twitter Sentiment Classification and Quantification∗ David Vilaresa, Yerai Dovala,b, Miguel A. Alonsoa and Carlos G´omez-Rodr´ıgueza aGrupo LyS, Departamento de Computaci´on, Universidade da Coru˜na Campus de A Coru˜na s/n, 15071, A Coru˜na, Spain bGrupo COLE, Departamento de Inform´atica, E.S. de Enxe˜nar´ıa Inform´atica, Universidade de Vigo Campus As Lagoas, 32004, Ourense, Spain {david.vilares, yerai.doval, miguel.alonso, carlos.gomez}@udc.es Abstract In this paper we describe our deep learning ap- proach for solving both two-, three- and five- class tweet polarity classification, and two- and five-class quantification. We first trained a convolutional neural network using pretrained Twitter word embeddings, so that we could extract the hidden activation values from the hidden layers once some input had been fed to the network. These values were then used as features for a support vector machine in both the classification and quantification subtasks, together with additional linguistic information in the former scenario. The results obtained for the classification subtasks show that this approach performs better than a single con- volutional network, and for the quantification part it also yields good results. Official rank- ings locate us: 2nd (practically tied with 1st) for the binary classification task, 2nd for bi- nary quantification and 4th (practically tied with 3rd) for the five-class polarity classifica- tion challenge. 1 Introduction Opinion mining has become an important mecha- nism to monitor what people are saying about a va- riety of topics (Cambria et al., 2013). As an ex- ample, Thelwall et al. (2011) use SentiStrength to ∗This research is supported by the Ministerio de Econom´ıa y Competitividad (FFI2014-51978-C2). David Vilares is funded by the Ministerio de Educaci´on, Cultura y Deporte (FPU13/01180). Yerai Doval is funded by the Ministerio de Econom´ıa y Competitividad (BES-2015-073768). Carlos G´omez-Rodr´ıguez is funded by an Oportunius program grant (Xunta de Galicia). monitor popular events on Twitter (e.g. the Oscars, the SuperBowl), showing how these resonate among the public. In a similar line, Vilares et al. (2015) also use Twitter to measure the level of popularity of the main Spanish political leaders, proving that the results obtained by their systems are similar to the ones obtained by traditional polls. Opinion mining on Twitter actually involves two different challenges: (1) analyzing the characteris- tics of individual tweets and (2) quantifying a given set of tweets so that useful statistics can be extracted. This paper describes our different models to over- come such challenges, using the SemEval 2016:Task 4 as the evaluation framework. 2 Sentiment Analysis in Twitter The SemEval organization proposed two different types of challenges in its 2016 edition: (1) classifi- cation into two, three and five classes and (2) quan- tification into two and five categories. A detailed explanation of the task can be found in the descrip- tion paper (Nakov et al., 2016). For all subtasks, three official splits are provided: training, develop- ment and development test sets. In this paper, we use …

Content truncated to 500 words