Title: Extracting customer opinions associated with an aspect by using a heuristic based sentence segmentation approach

Authors: Naime F. Kayaalp; Gary R. Weckman; William A. Young II; David Millie; Can Celikbilek

Addresses: Industrial and Systems Engineering, Russ College of Engineering and Technology, Ohio University, Athens, OH, 45701, USA ' Industrial and Systems Engineering, Russ College of Engineering and Technology, Ohio University, Athens, OH, 45701, USA ' Management Information Systems, College of Business, Ohio University, Athens, OH, 45701, USA ' Palm Island Enviro-Informatics, LLC Sarasota, FL, 34232, USA ' Industrial and Systems Engineering, Russ College of Engineering and Technology, Ohio University, Athens, OH, 45701, USA

Abstract: Extracting opinions from textual customer reviews is one of the most challenging but crucial tasks, which has a significant effect on decision support systems. The task becomes trickier when customers mention different aspects of an entity and different opinions associated with those aspects within the same review text. In literature, there are segmentation-based approaches, which try to parse the reviews based on the features/aspects they mention and assign opinions to those segments. It is observed that the existing models can perform better with an improved underlying segmentation model. This research aims to propose an improved segmentation model, named IMAS, to fill the gap in an existing multi-aspect segmentation model (MAS). Two aspect-opinion extraction systems were developed with two different segmentation models, and they are compared based upon their overall accuracy. As a result of the comparisons, an 8% accuracy improvement is observed by using the proposed segmentation model over the legacy one.

Keywords: review mining; aspect-opinion matching; text segmentation; decision support systems.

DOI: 10.1504/IJBIS.2017.086335

International Journal of Business Information Systems, 2017 Vol.26 No.2, pp.236 - 260

Received: 15 Jan 2016
Accepted: 06 May 2016
Published online: 04 Sep 2017 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article