Abstract
E-commerce is rapidly growing, with review Web sites hosting hundreds of reviews on average for any product. Reading so many reviews is tedious, time-consuming, and with the proposed Gist, unnecessary. We introduce Gist, a system to automatically summarize large amounts of text into informative and actionable key sentences. With unsupervised learning and sentiment analysis, Gist selects the sentences that best characterize a set of reviews. All of this is done in seconds, without prior adjustment or training. Gist extends the current state of the art with a modular system that can take advantage of a priori knowledge and adapt to new domains through easy modification and extension. Gist is a general framework, able to summarize any set of text and easily adapt to specific domains. A robust comparison with state-of-the-art summarization algorithms, on datasets containing hundreds of documents, proves Gist's ability to effectively summarize text and reviews.