Freely accessible and easily customizable Web applications have made it easy and fun for people to publish and share their experiences, knowledge, opinions, and emotions. The opinions of lay users (nonexperts) may serve to balance and complement the authoritative points of view published by mainstream media such as the New York Times newspaper and CNN; and the customer moods may help make market predictions. Therefore, user generated opinions have attracted an increasing amount of interest from individuals as well as organizations. For example, people are curious about what other people think of certain products or topics; companies want to find out what their target audience likes or dislikes about their products and services;
and government officials would like to learn whether people are for or against new policies. Foreseeing the growing demand for detecting online opinion, researchers from different communities started to explore a new research area called opinion mining in the late 1990s. Dave, Lawrence, and Pennock , who first coined the term “opinion mining,” described an opinion mining tool for online product reviews that aimed to automate the sequence of “processing a set of search results for a given item, generating a list of product attributes (quality, features, etc.) and aggregating opinions about each of them (poor, mixed, good)” (p. 519). To date, a series of opinion mining tasks have been explored, including differentiating opinions from facts, also known as subjectivity analysis [719, 722, 744, 767], detecting positive or negative opinion polarity [3, 156, 345, 371, 521], determining opinion strength [672, 724], and identifying opinion holders and opinion properties [71, 346, 371, 402, 616].