Sentiment Analysis in Twitter for Macedonian

Proceedings of Recent Advances in Natural Language Processing, pages 249–257, Hissar, Bulgaria, Sep 7–9 2015. Sentiment Analysis in Twitter for Macedonian Dame Jovanoski, Veno Pachovski University American College Skopje UACS, Macedonia {jovanoski,pachovski}@uacs.edu.mk Preslav Nakov Qatar Computing Research Institute HBKU, Qatar pnakov@qf.org.qa Abstract We present work on sentiment analysis in Twitter for Macedonian. As this is pio- neering work for this combination of lan- guage and genre, we created suitable re- sources for training and evaluating a sys- tem for sentiment analysis of Macedonian tweets. In particular, we developed a cor- pus of tweets annotated with tweet-level sentiment polarity (positive, negative, and neutral), as well as with phrase-level sen- timent, which we made freely available for research purposes. We further boot- strapped several large-scale sentiment lex- icons for Macedonian, motivated by pre- vious work for English. The impact of several different pre-processing steps as well as of various features is shown in ex- periments that represent the first attempt to build a system for sentiment analysis in Twitter for the morphologically rich Macedonian language. Overall, our exper- imental results show an F1-score of 92.16, which is very strong and is on par with the best results for English, which were achieved in recent SemEval competitions. 1 Introduction The increasing popularity of social media services such as Facebook, Twitter and Google+, and the advance of Web 2.0 have enabled users to share information and, as a result, to have influence on the content distributed via these services. The ease of sharing, e.g., directly from a laptop, a tablet or a smart phone, have contributed to the tremendous growth of the content that users share on a daily basis, to the extent that nowadays social networks have no choice but to filter part of the informa- tion stream even when it comes from our closest friends. Naturally, soon this unprecedented abundance of data has attracted business and research interest from various fields including marketing, political science, and social studies, among many others, which are interested in questions like these: Do people like the new Apple Watch? What do they hate about iPhone6? Do Americans support Oba- maCare? What do Europeans think of Pope’s visit to Palestine? How do we recognize the emergence of health problems such as depression? Such questions can be answered by studying the sentiment of the opinions people express in social media. As a result, the interest for sentiment anal- ysis, especially in social media, has grown, further boosted by the needs of various applications such as mining opinions from product reviews, detect- ing inappropriate content, and many others. Below we describe the creation of data and the development of a system for sentiment polarity classification in Twitter for Macedonian: positive, negative, neutral. We are inspired by a similar task at SemEval, which is an ongoing series of evalua- tions of computational semantic analysis systems, composed by multiple challenges such as text sim- ilarity, word sense disambiguation, etc. One of the challenges there was on Sentiment Analysis in Twitter, at …

Content truncated to 500 words