Detecting Fake Review with Rumor Model---Case Study in Hotel Review

• Published in: Intelligence Science and Big Data Engineering. Big Data and Machine Learning Techniques Vol. 9243; pp. 181 - 192
• Main Authors: Chang, Tien, Hsu, Ping Yu, Cheng, Ming Shien, Chung, Chen Yao, Chung, Yi Liang
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 01.01.2015; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Fake review; Graphics programming; Hotel review; Image processing; Pattern recognition; Rumor; Text mining
• ISBN: 3319238612; 9783319238616
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-23862-3_18

With the development of the Internet economy, various websites accumulate numerous reviews about different products and services. Those reviews have become one major information source besides official product information, expert opinion, and automatically generated individualized advice. The survey shows that percentage of gathering buying information on Internet gradually increases by years, and the relevant researchers have also proven that consumers pay more attention to others’ reviews, thus deeply affect consumers’ shopping decision. Unfortunately, by taking advantage of such trend, some dealers manipulate reviews in order to exaggerate their own product or defame their rivals. Those behaviors have brought severe damage to consumers and commerce. This study takes Internet reviews as research object, using rumor model to detect the truth of these review. Our rumor model applied text mining technique and extract 3 major characteristic of review content: important attribute word, specific quantifier, and noun verb ratio to build the model. For testing our rumor model, we take hotel reviews on America website “TripAdvisor” and the comparison group “Fake reviews” as analysis objects. We try to automatically and easily classify true and fake reviews. The result, generated by developed model in this research, shows that the more unique vocabulary and specific quantifier and noun it contains, the less possibility it is fake.