Filtering and Transformation Model for Opinion Summarization

Abstract

The rapid evolution of Micro blogging sites such as Blogs & Twitter facilitate people to post real time messages about their opinions on a variety of topics inclusive of products they use in their daily life. Summarizing opinions of bloggers has several interesting and commercially significant applications like helping the customer to reach purchasing decisions and as a guide for the business activities of companies such as product improvement & market adoption.The paper explores the data characteristics’ of Tweets/ Reviews which can be centrepiece of a conversation & provide excellent channel for opinion creation. The short length of the messages and their noisy nature makes it difficult to mine the micro blog data for opinions. Also the infrequent entities such as people, organization, products etc. and user creativity followed by freedom of language hinder the task of Opinion summarization. The paper demonstrates the major role played by Filtering and Transformation techniques in choosing representative words which is the basis for Features extraction in the task of Opinion summarization.

The paper concludes by proposing a framework for pre-processing which emphasises that feature reduction is an important step in Feature based summarization while not compromising on accuracy.