Feature Extraction from Online User Reviews
Sachith Paramie Karunathilake, Jayasekara Liyanapatabedige Achira Jeewaka Shamal, Rankothge Gishan Hiranya Pemathilake, Gamage Upeksha Ganegoda · 2018
Online reviews have a huge commercial value if analyzed correctly. Most common way of analyzing reviews is subjecting them to sentimental analysis. However, this does not provide fine grain actionable information due to the lack of specificity. This study suggests a mechanism to extract product features that have been discussed within the reviews there by allowing a more meaningful opinion analysis. This can be challenging as different features are expressed in different ways in reviews. The proposed method intends to use double propagation and rule-based mining to overcome the extraction problem. This feature extraction module is developed as a part of a larger online review analysis platform for electronic products which is currently being implemented. Finally, this paper presents the results observed in the experiments carried out on multiple datasets.