Opinion mining with reviews summarization based on clustering

Abstract

Automatic text summarization can be used in recommendation systems to present useful texts obtained from the available comments and texts. For summarization, a human reads all of the writing and gains a background understanding of the text, but computers do differently. Several methods have been proposed for automatic text summarization until now, from abstract summarization methods that deal with new sentences produced from important points existed in the texts to extraction summarization methods, which deal with original main sentences from the text. In this study, we present an extraction method for text summarizing. In this method, at first, the sentences are processed, and the similarities between sentences are calculated by a proposed similarity measure. Afterward, the sentences are clustered based on the similarities, and at last, a certain number of sentences are selected from each cluster. The Gaussian Mixture Model (GMM) algorithm is used to cluster the sentences. The proposed method is tested on a collected dataset from Tripadvisor ( https://www.tripadvisor.com/ ) customer reviews, and the results show that using GMM results in a more informative summary and more variation in sentences compared to K-means.