USF: Chunking for Aspect-term Identification & Polarity Classification

This paper describes the systems submitted by the University of San Francisco (USF) to Semeval-2014 Task 4, Aspect Based Sentiment Analysis (ABSA), which provides labeled data in two domains, laptops and restaurants. For the constrained condition of both the aspect term extraction and aspect term polarity tasks, we take a supervised machine learning approach using a combination of lexical, syntactic, and baseline sentiment features. Our extraction approach is inspired by a chunking approach, based on its strong past results on related tasks. Our system performed slightly below average compared to other submissions, possibly because we use a simpler classification model than prior work. Our polarity labeling approach uses two baseline hand-built sentiment classifiers as features in addition to lexical and syntactic features, and performed in the top ten of other constrained systems on both domains.