Opinion extraction from customer reviews

Abstract

The internet offers a new channel for product designers to obtain valuable information about customer's opinions which are very important to product development, especially at the product concept design stage. Due to the rapid growth of such information, it is difficult for humans to manage and analyze all these information. Therefore, an alternative choice is to perform opinion mining with automatic textual mining techniques. In this research, we propose a hybrid opinion extraction (HOE) framework that can extract features and predict semantic orientation of the expressed opinions, from the free format text. The framework is inspired by capturing the characteristics of the way people express opinions, utilizes both statistical regularities of the patterns and some prior knowledge. Compared to previous work, our opinion mining technique has demonstrated its better performance in terms of extracting features and predicting semantic orientations of opinions. Thus it has the potential to be adopted by product designers as an efficient tool for quickly obtaining customer feedback. Copyright © 2009 by ASME.