Exploring sentiment summarization

Abstract

We introduce the idea of a sentiment summary, a single passage from a document that captures a key aspect of the author's opinion about his or her subject. Using supervised data from the Rotten Tomatoes website, we examine features that appear to be helpful in locating a good summary sentence. These features are used to fit Naive Bayes and regularized logistic regression models for summary extraction. Copyright © 2004, American Association for Artificial Intelligence (www.aaai.org). All rights reserved.