Sentiment Analysis For Sports Fanaticism In Arabic Social Media Text
Husni Al-Muhtasab , Habib Rabaan, محمد مجاهد يحيى القماسي
كلية علوم وهندسة الحاسب الآلي-جامعة الملك فهد للبترول والمعادن · السعودية
Sport fanaticism is one of the social problems that have negative impact on social fabric. It is a psychological emotion that carries a blindness hatred against the competitive (teams, players, etc.) combined with a blindness love toward favorite (teams, players, etc.) where the emotional attitudes overcome the mental attitudes. Social media sites (e.g. Twitter) have become common means for communication between sport fans. Therefore, understanding how social media text contributes to increase or decrease sports-fanaticism is essential. We need to formulate the concept of Sports-Fanaticism into suitable criteria, rules, attributes, aspects, features and indicators that allow applying Arabic sentiment analysis techniques to help in automatically detecting and measuring this phenomenon in social network sites. The target of this research work is to automatically classify Arabic texts under consideration into fanatic and non-fanatic emotion. This will help the interested researchers and specialists in detecting and measuring the degree of sport fanaticism automatically. After formulating the problem, the proposed formalism was used as a guide to build annotated corpora and fanaticism-lexicons. To build these resources, around 4 million tweets were collected. We have used Positive Pointwise Mutual Information (PPMI) and Term-Frequency Inverse-Context-Frequency (TFICF) methods to generated ten fanatic-lexicons where Our proposed TFICF method showed better performance than PPMI in all experiments. The highest accuracy of TFICF is 91% and the highest accuracy of PPMI is 86%. Large-scale annotated corpora were also constructed. After building the required resources, we have conducted various experiments using different approaches, algorithms, and features. SVM, LR and NB classifiers were investigated. As one of the results, we have developed a classification tool that automatically evaluates a given text against sport fanaticism. such tool will help news sites to automatically filter messages posted by visitors to avoid sport fanaticism.Keywords: Sport Prejudice, Sport Intolerance, Opinion Mining, Social Networks, Machine Learning, Arabic Processing, Text Mining