抄録
This paper presents a computational method for discriminating travel reviews from commercial speech. Today many travel reviews can be found on the Web. Mining useful information from these reviews is important not only for consumers but also for companies like travel agencies. Travel reviews written by ordinary people are essentially different from commercial speech like advertisements generated by companies and individuals for the intent of making a profit. We propose a computational method for discriminating travel reviews from commercial speech. Assuming that subjective words often occur in travel reviews rather than commercial speech texts, we define the subjectivity score of each word in a document. Evaluating the total subjectivity score of a document using all the words in the document, the proposed method identifies whether the document is classified as travel review or not. From experiments, we have confirmed the proposed method can accurately classify documents into travel reviews and commercial speech texts.
