Feature subsumption for sentiment classification in multiple languages

摘要: An open problem in machine learning-based sentiment classification is how to extract complex features that outperform simple features; figuring out which types of are most valuable another Most the studies focus primarily on character or word Ngrams features, but substring-group have never been considered area before In this study, extracted and selected for by means transductive algorithm To demonstrate generality, experiments conducted three datasets different languages: Chinese, English Spanish The experimental results show proposed algorithm's performance usually superior best related work, feature subsumption multilingual Compared inductive algorithm, also illustrate can significantly improve As term weighting, “tfidf-c” outperforms all other weighting approaches algorithm.