MIS 373 Lecture Notes - Lecture 12: Pagerank, Sentiment Analysis
Document Summary
Useful when not accompanied by sentiment bearing adjectives or phrases. A more practical approach to unsupervised sentiment analysis. Can be made context sensitive by adjusting weights and/or adding words. Assumption: people express emotions in close proximity to the mentions of entities. Punctuation, e. g. , ! increases the sentiment intensity. Degree modifiers, e. g. , the service here is extremely good . Sentiment shifters, e. g. , the food here is great, but the service is horrible . Trigram preceding a sentiment-laden lexical feature: catch nearly 90% of cases where. E. g. , the food here isn"t really all that great . negation flips the polarity of the text. The b&w p7 are high on my favorite list next to the h8 by b&o and the grados. i also. Each product review (as discussed in the article) Primarily focuses on a product (say . Developed by larry page and serge brin at stanford. A variation of the good old eigenvector algebra.