Evaluating Distant Supervision for Subjectivity and Sentiment Analysis on Arabic Twitter Feeds

Abstract

Supervised machine learning methods for automatic subjectivity and sentiment analysis (SSA) are problematic when applied to social media, such as Twitter, since they do not generalise well to unseen topics. A possible remedy of this problem is to apply distant supervision (DS) approaches, which learn from large amounts of automatically annotated data. This research empirically evaluates the performance of DS approaches for SSA on Arabic Twitter feeds. Results for emoticon-and lexiconbased DS show a significant performance gain over a fully supervised baseline, especially for detecting subjectivity, where we achieve 95.19% accuracy, which is a 48.47% absolute improvement over previous fully supervised results.