Sentiment Lexicons for Arabic Social Media

Sentiment Lexicons for Arabic Social Media Saif M. Mohammad1, Mohammad Salameh2, Svetlana Kiritchenko1 1National Research Council Canada, 2University of Alberta saif.mohammad@nrc-cnrc.gc.ca, msalameh@ualberta.ca, svetlana.kiritchenko@nrc-cnrc.gc.ca Abstract Existing Arabic sentiment lexicons have low coverage—only a few thousand entries. In this paper, we present several large sentiment lexicons that were automatically generated using two different methods: (1) by using distant supervision techniques on Arabic tweets, and (2) by translating English sentiment lexicons into Arabic using a freely available statistical machine translation system. We compare the usefulness of new and old sentiment lexicons in the downstream application of sentence-level sentiment analysis. Our baseline sentiment analysis system uses numerous surface form features. Nonetheless, the system benefits from using additional features drawn from sentiment lexicons. The best result is obtained using the automatically generated Dialectal Hashtag Lexicon and the Arabic translation of the NRC Emotion Lexicon (accuracy of 66.6%). Finally, we describe a qualitative study of the automatic translations of English sentiment lexicons into Arabic, which shows that about 88% of the automatically translated entries are valid for Arabic as well. Close to 10% of the invalid entries are the result of gross mistranslation, close to 40% are due to translation into a related word, and about 50% are due to differences in how the word is used in Arabic. Keywords: Arabic sentiment lexicons, sentiment analysis, social media, Arabic texts, valence, opinion, translation 1. Introduction Sentiment lexicons are lists of positive and negative words, optionally with a score indicating the degree of polarity. Sentiment analysis systems, including the best performing ones such as the NRC-Canada system (Mohammad et al., 2013; Kiritchenko et al., 2014a; Zhu et al., 2014), use sen- timent lexicons to obtain significant improvements (Wilson et al., 2013; Pontiki et al., 2014; Rosenthal et al., 2015; Mo- hammad et al., 2016a). However, much of the past work has focused on English texts and English sentiment lexicons. Arabic sentiment analysis has benefited from recent work (Farra et al., 2010; Abdul-Mageed et al., 2011; Badaro et al., 2014; Refaee and Rieser, 2014), but a resource that is particularly lacking is a large sentiment lexicon. Existing lexicons contain only a few hundred to a few thousand en- tries. Further, most do not indicate the degree of association between a word and positive (or negative) sentiment. Kir- itchenko et al. (2016) created a manually annotated Arabic sentiment lexicon with real-valued sentiment scores; how- ever, that lexicon too has only around one thousand entries. In this paper, we describe how we created Arabic senti- ment lexicons with tens of thousands of entries. These in- clude new automatically generated ones as well as transla- tions of existing English lexicons. We use the lexicons (old and new) for sentiment classification of Arabic social me- dia posts. Our baseline system uses numerous surface form features. Nonetheless, the system benefits from using ad- ditional features drawn from sentiment lexicons. We also show the extent to which various sentiment lexicons are ef- fective in sentiment analysis. The best result was obtained using both the Dialectal …

Content truncated to 500 words