Paper Type
Completed Research Paper
Abstract
Along with the exponential growth of user-generated content online comes the need of making sense of such content. Online consumer review is one type of user-generated content that has been more important. Thus, there is a demand for uncovering hidden patterns, unknown relationships and other useful information. The focal problem of this research is product feature extraction. Few existing studies has looked into detailed categorization of review features and explored how to adjust extraction methods by taking account of the characteristics of different categories of features. This paper begins with the introduction of a new scheme of feature classification and then introduces new extraction methods for each type of features separately. These methods were design to not only recognize new features but also filter irrelevant features. The experimental results show that our proposed methods outperform the state-of-the-art techniques.
Recommended Citation
Kang, Yin and Zhou, Lina, "Extracting Product Features from Online Consumer Reviews" (2013). AMCIS 2013 Proceedings. 2.
https://aisel.aisnet.org/amcis2013/IntelligentSystems/GeneralPresentations/2
Extracting Product Features from Online Consumer Reviews
Along with the exponential growth of user-generated content online comes the need of making sense of such content. Online consumer review is one type of user-generated content that has been more important. Thus, there is a demand for uncovering hidden patterns, unknown relationships and other useful information. The focal problem of this research is product feature extraction. Few existing studies has looked into detailed categorization of review features and explored how to adjust extraction methods by taking account of the characteristics of different categories of features. This paper begins with the introduction of a new scheme of feature classification and then introduces new extraction methods for each type of features separately. These methods were design to not only recognize new features but also filter irrelevant features. The experimental results show that our proposed methods outperform the state-of-the-art techniques.