A Structure for Opinion in Social Domains

摘要

Opinion mining, as a sub-field of text mining, analyzes opinions expressed regarding an object, a topic, or an issue. An opinion is expressed by a person using some opinion terms or phrases regarding a target. Statistical studies show that the affective factors on opinions in product domains are different from those in social domains. Opinion verbs and ``I" have the most affective influences on opinions in social domains. This paper introduces a structure for opinions in social domains considering verb as its core. An outline for opinion extraction from text is also proposed. The defined structure is evaluated in the applications of sentence subjectivity classification and sentiment polarity classification at the sentence and document levels. Our experiments show that the performance of the proposed structure is slightly higher than the traditional machine learning techniques and some previous works.