An Architecture for Automatic Opinion Classification in Western Online News

Drawing on studies in media discourse analysis and advances in natural language processing, data mining and information access, a framework for automated sentiment analysis of news report from Western mainstream media is proposed and a preliminary experiment is reported in this paper. The automated sentiment analysis system architecture comprises (1) the identification of Chinese-English equivalent query expressions; (2) the construction of domain and linguistic knowledge bases; (3) the recovery of news events; and (4) the sentiment analysis over news texts. The workflow consists of two phases, i.e. an offline construction of resources, such as comparable or parallel corpora and domain knowledge bases, which are utilized in real-time processing, and an online analysis of news texts by means of machine learning and data mining techniques. The proposed methodology leads to a quick response to different information gathered from overseas media and an efficient decision-making procedure for e-government.