Proceedings of the Workshop on Sentiment and Subjectivity in Text

This volume contains the papers prepared for and presented at the Workshop on Sentiment and Subjectivity in Text, held on 22 July 2006 in Sydney, Australia, immediately following COLING/ACL 2006. Sentiment and subjectivity in text constitute a problem that is orthogonal to typical topic detection tasks in text classification. Despite the lack of a precise definition of sentiment or subjectivity, headway has been made in matching human judgments by automatic means. Such systems can prove useful in a variety of contexts. In many applications it is important to distinguish what an author is talking about from his or her subjective stance towards the topic. If the writing is highly subjective, as for example in an editorial text or comment, the text should be treated differently than if it were a mostly objective presentation of facts, as for example in a news article. Information extraction, summarization, and question answering can benefit from an accurate separation of subjective content from objective content. Furthermore, the particular sentiment expressed by an author towards a topic is important for "opinion mining", i.e. the extraction of prevalent opinions about topics or items from a collection of texts. Similarly, in business intelligence it is important to automatically extract positive and negative perceptions about features of a product or service.