This paper introduces OPINE, an unsupervised information-extraction system for mining product reviews to identify important product features, their evaluation by reviewers, and their relative quality across products. OPINE achieves 22% higher precision (with only 3% lower recall) on the feature extraction task compared to previous work. Its novel use of relaxation labeling for finding the semantic orientation of words in context leads to strong performance in identifying opinion phrases and their polarity. OPINE decomposes the review mining problem into four subtasks: identifying product features, identifying opinions regarding product features, determining the polarity of opinions, and ranking opinions based on their strength. It is built on top of the KnowItAll Web information-extraction system and uses Web PMI statistics to improve feature extraction. OPINE's Feature Assessor evaluates candidate features using PMI scores, and its relaxation labeling technique determines the semantic orientation of potential opinion words in the context of given features and sentences. OPINE extracts explicit features from reviews and identifies corresponding customer opinions about these features, determining their polarity. It uses syntactic dependencies and relaxation labeling to extract opinion phrases and determine their polarity. OPINE's performance is evaluated on 7 product classes and compared to previous systems, showing higher precision and recall in opinion phrase extraction and polarity determination. OPINE's use of relaxation labeling allows it to identify customer opinions and their polarity with high precision and recall. The system is evaluated on tasks such as finding SO labels of words in the context of known features and sentences, distinguishing between opinion and non-opinion phrases, and finding the correct polarity of extracted opinion phrases. OPINE outperforms previous systems in these tasks, demonstrating its effectiveness in extracting and analyzing opinion phrases corresponding to specific features in specific sentences.This paper introduces OPINE, an unsupervised information-extraction system for mining product reviews to identify important product features, their evaluation by reviewers, and their relative quality across products. OPINE achieves 22% higher precision (with only 3% lower recall) on the feature extraction task compared to previous work. Its novel use of relaxation labeling for finding the semantic orientation of words in context leads to strong performance in identifying opinion phrases and their polarity. OPINE decomposes the review mining problem into four subtasks: identifying product features, identifying opinions regarding product features, determining the polarity of opinions, and ranking opinions based on their strength. It is built on top of the KnowItAll Web information-extraction system and uses Web PMI statistics to improve feature extraction. OPINE's Feature Assessor evaluates candidate features using PMI scores, and its relaxation labeling technique determines the semantic orientation of potential opinion words in the context of given features and sentences. OPINE extracts explicit features from reviews and identifies corresponding customer opinions about these features, determining their polarity. It uses syntactic dependencies and relaxation labeling to extract opinion phrases and determine their polarity. OPINE's performance is evaluated on 7 product classes and compared to previous systems, showing higher precision and recall in opinion phrase extraction and polarity determination. OPINE's use of relaxation labeling allows it to identify customer opinions and their polarity with high precision and recall. The system is evaluated on tasks such as finding SO labels of words in the context of known features and sentences, distinguishing between opinion and non-opinion phrases, and finding the correct polarity of extracted opinion phrases. OPINE outperforms previous systems in these tasks, demonstrating its effectiveness in extracting and analyzing opinion phrases corresponding to specific features in specific sentences.