OMAM at SemEval-2017 Task 4: Evaluation of English State-of-the-Art Sentiment Analysis Models for Arabic and a New Topic-based Model

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 603–610, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics OMAM at SemEval-2017 Task 4: Evaluation of English State-of-the-Art Sentiment Analysis Models for Arabic and a New Topic-based Model Ramy Baly1, Gilbert Badaro1, Ali Hamdi2 Rawan Moukalled1, Rita Aoun1, Georges El-Khoury1, Ahmad El-Sallab3 Hazem Hajj1, Nizar Habash4, Khaled Bashir Shaban5, Wassim El-Hajj6 1 Department of Electrical and Computer Engineering, American University of Beirut 2 Faculty of Computing, Universiti Teknologi Malaysia 3 Computer Engineering Department, Cairo University 4 Computational Approaches to Modeling Language Lab, New York University Abu Dhabi 5 Computer Science and Engineering Department, Qatar University 6 Department of Computer Science, American University of Beirut {rgb15,ggb05,rrm32,rra47,gbe03,hh63,we07}@mail.aub.edu, nizar.habash@nyu.edu, ali@alihamdi.com ahmad.elsallab@gmail.com, khaled.shaban@qu.edu.qa Abstract While sentiment analysis in English has achieved significant progress, it remains a challenging task in Arabic given the rich morphology of the language. It becomes more challenging when applied to Twitter data that comes with additional sources of noise including dialects, misspellings, grammatical mistakes, code switching and the use of non-textual objects to ex- press sentiments. This paper describes the “OMAM” systems that we developed as part of SemEval-2017 task 4. We eval- uate English state-of-the-art methods on Arabic tweets for subtask A. As for the remaining subtasks, we introduce a topic- based approach that accounts for topic specificities by predicting topics or do- mains of upcoming tweets, and then us- ing this information to predict their senti- ment. Results indicate that applying the English state-of-the-art method to Ara- bic has achieved solid results without sig- nificant enhancements. Furthermore, the topic-based method ranked 1st in subtasks C and E, and 2nd in subtask D. 1 Introduction Sentiment Analysis (SA) is a fundamental prob- lem aiming to allow machines to automatically ex- tract subjectivity information from text (Turney, 2002), whether at the sentence or the document level (Farra et al., 2010). This field has been attracting attention in the research and business communities due to the complexity of human lan- guage, and given the range of applications that are interested in harvesting public opinion in different domains such as politics, stocks and marketing. The interest in SA from Arabic tweets has in- creased since Arabic has become a key source of the Internet content (Miniwatts, 2016), with Twitter being one of the most expressive social media platforms. While models for SA from English tweets have achieved significant success, Arabic methods continue to lag. Opinion mining in Arabic (OMA) is a challenging task given: (1) the morphological complexity of Arabic (Habash, 2010), (2) the excessive use of dialects that vary significantly across the Arab world, (3) the sig- nificant amounts of misspellings and grammati- cal errors due to length restriction in Twitter, (4) the variations in writing styles, topics and expres- sions used across the Arab world due to cultural diversity (Baly et al., 2017), and (5) the exis- tence of Twitter-specific tokens (hashtags, men- tions, multimedia objects) that may have subjec- tive information embedded in them. …

Content truncated to 500 words