Effects of Adjective Orientation and Gradability on Sentence Subjectivity

Effects of Adjective Orientation and Gradability on Sentence Subjectivity Vasileios Hatzivassiloglou Department of Computer Science Columbia University New York, NY 10027 vh@cs.columbia.edu Janyce M. Wiebe Department of Computer Science New Mexico State University Las Cruces, NM 88003 wiebe@cs.nmsu.edu Abstract Subjectivity is a pragmatic, sentence-level feature that has important implications for text processing applica- tions such as information extraction and information re- trieval. We study the effects of dynamic adjectives, se- mantically oriented adjectives, and gradable adjectives on a simple subjectivity classifier, and establish that they are strong predictors of subjectivity. A novel train- able method that statistically combines two indicators of gradability is presented and evaluated, complementing existing automatic techniques for assigning orientation labels. 1 Introduction In recent years, computational techniques for the deter- mination of lexical semantic features have been proposed and evaluated. Such features include sense, register, do- main specificity, pragmatic restrictions on usage, seman- tic markedness, and orientation, as well as automatically identified links between words (e.g., semantic related- ness, synonymy, antonymy, and meronymy). Automat- ically learning features of this type from large corpora allows the construction or augmentation of lexicons, and the assignment of semantic labels to words and phrases in running text. This information in turn can be used to help determine additional features at the lexical, clause, sentence, or document level. This paper explores the benefits that some lexical fea- tures of adjectives offer for the prediction of a contextual sentence-level feature, subjectivity. Subjectivity in nat- ural language refers to aspects of language used to ex- press opinions and evaluations. The computational task addressed here is to distinguishsentences used to present opinions and other forms of subjectivity (subjective sen- tences, e.g., “At several different layers, it’s a fascinating tale”) from sentences used to objectively present factual information (objective sentences, e.g., “Bell industries Inc. increased its quarterly to 10 cents from 7 cents a share”). Much research in discourse processing has focused on task-oriented and instructional dialogs. The task ad- dressed here comes to the fore in other genres, especially news reporting and Internet forums, in which opinions of various agents are expressed and where subjectivity judgements could help in recognizing inflammatory mes- sages (“flames”) and mining online sources for product reviews. Other tasks for which subjectivity recognition is potentially very useful include information extraction and information retrieval. Assigning subjectivity labels to documents or portions of documents is an example of non-topical characterization of information. Current in- formation extraction and retrieval technology focuses al- most exclusively on the subject matter of the documents. Yet, additional components of a document influence its relevance to particular users or tasks, including, for ex- ample, the evidential status of the material presented, and attitudes adopted in favor or against a particular person, event, or position (e.g., articles on a presidential cam- paign written to promote a specific candidate). In sum- marization, subjectivity judgments could be included in document profiles to augment automatically produced document summaries, and to help the user make rele- vance judgments when using a search engine. …

Content truncated to 500 words