This paper presents a holistic lexicon-based approach to opinion mining, focusing on determining the semantic orientation (positive, negative, or neutral) of opinions expressed on product features in customer reviews. Existing methods rely on opinion lexicons, which list words expressing positive or negative sentiments. However, these methods struggle with context-dependent opinions and conflicting opinions in a sentence. The proposed approach addresses these challenges by incorporating external evidence, linguistic conventions, and context to infer the semantic orientation of opinion words. It also handles special words, phrases, and language constructs that influence opinions based on their linguistic patterns. The system, called Opinion Observer, has been implemented and tested on benchmark product review data sets, showing significant improvements over existing methods. The approach involves identifying opinion words, phrases, and idioms, and using a score function that considers the distance between opinion words and product features to determine the semantic orientation. It also handles negation, conjunction, and context-dependent opinions by applying linguistic rules and using information from other sentences or reviews. The method is effective in aggregating conflicting opinions and handling implicit features, which are not always explicitly mentioned. The proposed technique outperforms existing methods in terms of precision, recall, and F-score. It is able to handle context-dependent opinions and aggregate multiple conflicting opinions in a sentence, making it more robust and accurate. The results show that the new method is highly effective and significantly better than existing approaches in opinion mining.This paper presents a holistic lexicon-based approach to opinion mining, focusing on determining the semantic orientation (positive, negative, or neutral) of opinions expressed on product features in customer reviews. Existing methods rely on opinion lexicons, which list words expressing positive or negative sentiments. However, these methods struggle with context-dependent opinions and conflicting opinions in a sentence. The proposed approach addresses these challenges by incorporating external evidence, linguistic conventions, and context to infer the semantic orientation of opinion words. It also handles special words, phrases, and language constructs that influence opinions based on their linguistic patterns. The system, called Opinion Observer, has been implemented and tested on benchmark product review data sets, showing significant improvements over existing methods. The approach involves identifying opinion words, phrases, and idioms, and using a score function that considers the distance between opinion words and product features to determine the semantic orientation. It also handles negation, conjunction, and context-dependent opinions by applying linguistic rules and using information from other sentences or reviews. The method is effective in aggregating conflicting opinions and handling implicit features, which are not always explicitly mentioned. The proposed technique outperforms existing methods in terms of precision, recall, and F-score. It is able to handle context-dependent opinions and aggregate multiple conflicting opinions in a sentence, making it more robust and accurate. The results show that the new method is highly effective and significantly better than existing approaches in opinion mining.