Posted on Aug. 23, 2014
Last spring, the LT3 team participated in the Semantic Evaluation challenge on Sentiment Analysis in Twitter. The challenge consisted of both contextual polarity disambiguation (subtask A) and message polarity classification (subtask B). Our tool, named LT3, differentiated itself by performing feature selection on a wide range of features (lexical, syntactic, semantic and twitter-specific). We ranked among the first 13 out of 27 submissions for subtask A, and among the first 18 out of 42 submissions for subtask B. This work is being presented the next two days at Coling 2014 in Dublin, Ireland.