Identifying and classifying subjective claims

To understand the subjective documents, for example, public comments on the government’s proposed regulation, opinion identification and classification is required. Rather than classifying documents or sentences into binary polarities as in much previous work, we identify the main claim or assertion of the writer and classify it into the predefined classes of opinion (attitude) over the topic. For the classification of the claims, we automatically build a list of multi-word subjective expressions by extending a small set of seed words, using automatically generated paraphrases from machine translation corpus. Our supervised machine learning method shows significant improvement over the baseline both in identification and classification of claims. Categories and Subject Descriptors I.2.7 [Natural Languages]: Text analysis General Terms Algorithms, Experimentation, Languages Keywords Electronic Rulemaking, Public Comments, Text Annotation, Opinion

• Type: Conference
• Year: 2007
• Where: DGO
• Authors: Namhee Kwon, Liang Zhou, Eduard H. Hovy, Stuart W. Shulman
• Keywords: DGO 2007, Education, Government’s Proposed Regulation, Multi-word Subjective Expressions, Public Comments