Enhancing the performance of sentiment analysis task on product reviews by handling both local and global context
Bagus Setya Rintyarna, Riyanarto Sarno, Chastine Fatichah
Abstract
Commonly, product review analysis includes extracting sentiment from product documents. The contextual aspect contained in a review document has potential to improve results obtained by the sentiment analysis task. In this regard, this paper proposes an approach that takes into account both local and global context. The main contribution of this work is threefold. Firstly, local context is defined and the graph-based word sense disambiguation (WSD) method is extended to assign the correct sense of a word in the context of a sentence. Secondly, global context is defined for addressing contextual issues related to the specific domain of a review document by using an improved SentiCircle-based method. Thirdly, a weighted mean-based strategy to determine sentiment value at document level is presented. Several experiments were conducted to assess the proposed method. Overall, the proposed method outperformed the baseline method in the metrics of precision, recall, F-measure and accuracy.