Semi-supervised Sentiment Classification in Resource-Scarce Language : A Comparative Study

概要

論文の詳細を見る
With the advent of consumer generated media (e.g., Amazon reviews, Twitter, etc.), sentiment classification becomes a heated topic. Conventional approaches heavily rely on a large amount of linguistic resources, which are difficult to obtain in resource-scarce languages. To overcome this problem, semi-supervised learning (SSL) algorithms have been exploited. However, for the development and variety involved in SSL literature, when people try to adopt SSL approach in practice, they usually confront difficulty in deciding the proper method from many potential candidates. In this study, we conduct empirical evaluation on several representative SSL algorithms in a document-level sentiment classification task for resource-scarce languages (Chinese in our case), and the comparative experiment is carried out using three real datasets. We will describe corresponding theorems, show characteristics and related existing issues for each evaluated algorithm. We believe the other people who interested in exploiting SSL methods could benefit from our experience.
2012-07-25