This paper introduces OPINE, an unsupervised information-extraction system designed to mine product reviews for important features, their evaluation by reviewers, and their relative quality across products. OPINE achieves 22% higher precision (with only 3% lower recall) on feature extraction compared to previous work. The system's novel use of relaxation labeling for finding the semantic orientation of words in context leads to strong performance on opinion phrase extraction and polarity determination. OPINE is built on the Know-It-All Web information-extraction system and is evaluated on various datasets, demonstrating robust performance across multiple product classes. The paper also discusses related work and concludes by highlighting OPINE's contributions and potential applications in generating opinion summaries.This paper introduces OPINE, an unsupervised information-extraction system designed to mine product reviews for important features, their evaluation by reviewers, and their relative quality across products. OPINE achieves 22% higher precision (with only 3% lower recall) on feature extraction compared to previous work. The system's novel use of relaxation labeling for finding the semantic orientation of words in context leads to strong performance on opinion phrase extraction and polarity determination. OPINE is built on the Know-It-All Web information-extraction system and is evaluated on various datasets, demonstrating robust performance across multiple product classes. The paper also discusses related work and concludes by highlighting OPINE's contributions and potential applications in generating opinion summaries.