Sentiment Classification for Web Search Results

Abstract

This study proposes an approach to display Google search results with different classes of sentimental orientations: (1) positive, negative, or neutral, (2) positive or negative, (3) positive or non-positive, and (4) negative or non-negative. A prototype, called as GSCS was also constructed to retrieve the search results of smartphones, tablets, and notebooks from Google. With a single click, the GSCS would help users easily get the opinions that they want to meet their different needs. For classifying documents, we suggest a two-level sentiment classification approach. At the sentence level, sentences are first classified into positive, negative, or neutral, and then the sentiment labels of the sentences were used in the classification of documents. We also demonstrated that our two-level sentiment classification (first sentence level and then document level) outperformed the document-level-only sentiment classification.