Learning to Analyze Relevancy and Polarity of Tweets.
Rianne Kaptein · 2012
Abstract. This paper describes the participation of Oxyme in the profiling task of the RepLab workshop. We use a machine learning approach to predict the relevancy and polarity for reputation. The same classifier is used for both tasks. Features used include query dependent features, relevancy features, tweet features and sentiment features. An important component of the relevancy features are manually provided positive and negative feedback terms. Our best run uses a Naive Bayes classifier and reaches an accuracy of 41.2 % on the profiling task. Relevancy of tweets is predicted with an accuracy of 80.9%. Predicting polarity for reputation turns out to be more difficult, the best polarity run achieves an accuracy of 38.1%. 1