SINAI at SemEval-2017 Task 4: User based classification

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 634–639, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics SINAI at SemEval-2017 Task 4: User based classification Salud Mar´ıa Jim´enez-Zafra, Arturo Montejo-R´aez, M. Teresa Mart´ın-Valdivia, L. Alfonso Ure˜na-L´opez Computer Science Department, Escuela Polit´ecnica Superior de Ja´en Universidad de Ja´en, 23071 - Ja´en (Spain) {sjzafra, amontejo, maite, laurena}@ujaen.es Abstract This document describes our participation in SemEval-2017 Task 4: Sentiment Anal- ysis in Twitter. We have only reported results for subtask B - English, deter- mining the polarity towards a topic on a two point scale (positive or negative senti- ment). Our main contribution is the inte- gration of user information in the classifi- cation process. A SVM model is trained with Word2Vec vectors from user’s tweets extracted from his timeline. The obtained results show that user-specific classifiers trained on tweets from user timeline can introduce noise as they are error prone be- cause they are classified by an imperfect system. This encourages us to explore further integration of user information for author-based Sentiment Analysis. 1 Introduction Task 4 of SemEval 2017, Sentiment Analysis in Twitter (Rosenthal et al., 2017), has included some new subtasks this year. One of these subtasks con- siders user information to be also integrated in pro- posed systems. We have participated in subtask B consisting of, given a message and a topic, clas- sify the message on a two-point scale (positive or negative sentiment towards that topic). Actually, organizers provide scripts to download user pro- file information such as age, location, followers... We have taken advantage of this information to ex- pand a SVM model trained with Word2Vec vec- tors from user publications on this social media. In this paper, we present our approach to clas- sify tweets in a two point scale (positive and negative) by combining Support Vector Machine (SVM), Word2Vec (Mikolov et al., 2013) and user information. We have decided to combine these technologies for several reasons. Firstly, we have applied SVM many different tasks includ- ing tweet polarity classification with good results (Saleh et al., 2011). Secondly, after a revision of the systems presented in the last year for the same task (Nakov et al., 2016), it seems that bet- ter results are achieved by using word embeddings representations, so we have decided to test how it works on user modeling. Finally, this year for the first time, organizers include user information. We consider that it is very interesting to integrate this contextual information to improve tweets sen- timent classification. Actually, polarity classifi- cation on a per-user basis has been found to be useful in tasks like collaborative filtering (Garc´ıa- Cumbreras et al., 2013). Besides, the generation of user profiles in Twitter has attracted the atten- tion of many researches in recent years, enabling the prediction of user behavior as in election pro- cesses (Pennacchiotti and Popescu, 2011). In Section 2 we explain the data used in our ap- proach. Section 3 presents the system description. Experiments and …

Content truncated to 500 words