Research On Aspect-based Summarization On Hotel Reviews Usingdouble Propagation And Fuzzy Linguistic Hedge

Historically,it is customary to refer to the opinion of a family member,a friend or an expert before making any decision on all aspects of life.This custom is always current and it's even more systematic with Internet more and more accessible to everyone and with all the social network steadily growing.Sites where you can find reviews of customers who used services or products are very legion.The hospitality domain is no exception to the rule.However,with this very invasive stream of review and so much information to manage it is very difficult or impossible to extract from its data useful information or even to make a quick analysis.Aspect-based sentiment Analysis(ASBA)may help to find relevant aspect and its polarity from a collection of hotel reviews.This paper aims to build aspect-based summarization system employing modified Double Propagation(DP)for extracting pairs of aspects and its sentiment.We implement a Fuzzy logic with linguistic hedge to perform a fine grain sentiment polarity.The use of Senti Word Net helped us to calculate the fuzzy membership of each opinion.We then aggregate each aspect-polarity pair to its relevant category and presenting them as a summary.DP propagation algorithm is modified to extract target-opinion word pair and linguistic hedge.We also add additional pruning method for extracted opinion word with DP and add methods to detect aspect phrase.Aspect categorization is conducted by creating an aspect-to-category corpus and calculating cosine similarity score between category's seed with aspect expression that will be categorized.The summary is presented by grouping aspect expressions of each aspect category based on its sentiment orientation.Based on evaluation,our approach is better than the existing state of the art approaches and can be applied to hotel review to present aspect-based summarization.The performance still needs some improvements in future works to handle verbs opinions,verb product features and subjective sentences.