This paper introduces a method for inferring the semantic orientation of a word from its statistical association with a set of positive and negative paradigm words. Semantic orientation refers to the evaluative character of a word, with positive orientation indicating praise and negative orientation indicating criticism. The method uses two statistical measures of word association: pointwise mutual information (PMI) and latent semantic analysis (LSA). The approach is tested on 3,596 words, including adjectives, adverbs, nouns, and verbs, with 1,614 positive and 1,982 negative words. The method achieves an accuracy of 82.8% on the full test set, but accuracy rises above 95% when the algorithm is allowed to abstain from classifying mild words. The paper discusses the applications of semantic orientation, including text classification, text filtering, tracking opinions in online discussions, analysis of survey responses, and automated chat systems. It also explores related work, such as the supervised learning algorithm developed by Hatzivassiloglou and McKeown, and the use of semantic orientation in review classification and sentiment analysis. The paper presents experimental results showing that the method performs well on different corpora and that the accuracy can be improved by adjusting parameters such as the Laplace smoothing factor and the neighborhood size. The results indicate that the method is effective for inferring semantic orientation from word association and has potential applications in various natural language processing tasks.This paper introduces a method for inferring the semantic orientation of a word from its statistical association with a set of positive and negative paradigm words. Semantic orientation refers to the evaluative character of a word, with positive orientation indicating praise and negative orientation indicating criticism. The method uses two statistical measures of word association: pointwise mutual information (PMI) and latent semantic analysis (LSA). The approach is tested on 3,596 words, including adjectives, adverbs, nouns, and verbs, with 1,614 positive and 1,982 negative words. The method achieves an accuracy of 82.8% on the full test set, but accuracy rises above 95% when the algorithm is allowed to abstain from classifying mild words. The paper discusses the applications of semantic orientation, including text classification, text filtering, tracking opinions in online discussions, analysis of survey responses, and automated chat systems. It also explores related work, such as the supervised learning algorithm developed by Hatzivassiloglou and McKeown, and the use of semantic orientation in review classification and sentiment analysis. The paper presents experimental results showing that the method performs well on different corpora and that the accuracy can be improved by adjusting parameters such as the Laplace smoothing factor and the neighborhood size. The results indicate that the method is effective for inferring semantic orientation from word association and has potential applications in various natural language processing tasks.