Mining overall sentiment in large sets of opinions

Abstract

Nowadays e-commerce becomes more and more popular and widespread. Being on the web opens many possibilities for both customers and service providers or goods merchants, thence there is a great potential and opportunity for all e-shops. For example, many e-shops provide discussion forum for every product they sell. Besides the primary role of any discussion forum, i.e. to facilitate discussions among customers, the reader can benefit from the opinions expressed there. However, if there are too many comments on the forum, we have a problem of extracting the essential overall opinion from it. At least some kind of partial automation is desirable. One approach is to perform opinion mining. In this work, we elaborate opinion mining for a problem setting where there is a need to extract overall sentiments expressed by discussion forum participants, i.e. whether the opinion represents positive or negative attitude. We propose a method of mining overall sentiments and design a fully automated system that will provide intelligent processing of large amount of reviews and opinions. © Springer-Verlag Berlin Heidelberg 2010.