EN-ES-CS: An English-Spanish Code-Switching Twitter Corpus for Multilingual Sentiment Analysis

EN-ES-CS: An English-Spanish Code-Switching Twitter Corpus for Multilingual Sentiment Analysis David Vilares, Miguel A. Alonso, Carlos G´omez-Rodr´ıguez Grupo LyS, Departamento de Computaci´on, Universidade da Coru˜na Campus de A Coru˜na s/n, 15071, A Coru˜na, Spain {david.vilares, miguel.alonso, carlos.gomez}@udc.es Abstract Code-switching texts are those that contain terms in two or more different languages, and they appear increasingly often in social media. The aim of this paper is to provide a resource to the research community to evaluate the performance of sentiment classification techniques on this complex multilingual environment, proposing an English-Spanish corpus of tweets with code-switching (EN-ES-CS CORPUS). The tweets are labeled according to two well-known criteria used for this purpose: SentiStrength and a trinary scale (positive, neutral and negative categories). Preliminary work on the resource is already done, providing a set of baselines for the research community. Keywords: Sentiment Analysis, Corpus Generation, Code-Switching 1. Introduction Sentiment analysis (SA) is the field of research that deals with the automatic comprehension of the subjective infor- mation shared by users, especially on the Web (Pang and Lee, 2008; Cambria et al., 2013). One of its main chal- lenges is polarity classification, focused on classifying a text, sentence or phrase as positive, negative or neutral (or even considering different levels of granularity like strongly positive and strongly negative (Nakov et al., 2016)). The interest of organizations and companies in this task has in- creased in recent years, due to the rise of social media. In particular, Twitter has become one of the most useful so- cial networks for trending analysis, thanks to its abilities to capture popular trends and the easy interaction among its members. SA techniques have been successfully applied on Twitter to monitor a wide variety of issues ranging from the per- ception of the public with respect to popular events (Thel- wall et al., 2011) to real-time political analysis (Vilares et al., 2015d). Some of these trends are global (e.g. the Oscars, Superbowl or Rihanna) and so their trending top- ics are global too (e.g. ‘#oscars2016’, ‘#superbowl2016’, ...). However, the public perception of these trends often changes from one country to another and the task becomes even harder when tweets are written in different languages. This has motivated the need of multilingual sentiment anal- ysis. Usually, researchers evaluate multilingual approaches by translating (Balahur and Turchi, 2012) or merging monolingual corpora in different languages (Vilares et al., 2015c). But there exist cases where these synthetic corpora are not adequate to evaluate more difficult and unexplored multilingual variants, such as code-switching texts (i.e. texts that contain terms in two or more different languages). Colloquial creole languages such as Spanglish (a mix of Spanish and American English) or Singlish (English-based creole from Singapore) or even official languages such as the Haitian creole (which merges Portuguese, Spanish, Ta´ıno, and West African languages), are some of the best- known situations of spoken code-switching. The aim of this paper is to provide a resource to the research community that can be used to evaluate the performance of sentiment …

Content truncated to 500 words