A faceted characterization of the opinion mining landscape

Increasing amounts of user generated content (UGC) on the Internet, is creating research interest in opinion mining. This involves automatic detection of opinions about products, services, political parties, celebrities and events from user generated content. Research efforts in opinion mining are thus far, fragmented and have followed several approaches. However, most of them require understanding of domain-specific opinion words and their polarity, and language-specific opinion rules. In addition, semantic constructs like sarcasm pose open challenges. In this paper, we try to characterize the opinion-mining landscape by proposing a faceted taxonomy of the different aspects of opinion mining. We then survey literature and place these in appropriate places in the proposed model. We also propose a general purpose workflow required from any opinion mining engine. Finally, we speculate on specific challenges in the opinion mining landscape.