Skip to Main content Skip to Navigation
Conference papers

Comparing Human Computation, Machine, and Hybrid Methods for Detecting Hotel Review Spam

Abstract : Most adults in industrialized countries now routinely check online reviews before selecting a product or service such as lodging. This reliance on online reviews can entice some hotel managers to pay for fraudulent reviews – either to boost their own property or to disparage their competitors. The detection of fraudulent reviews has been addressed by humans and by machine learning approaches yet remains a challenge. We conduct an empirical study in which we create fake reviews, merge them with verified reviews and then employ four methods (Naïve Bayes, SVMs, human computation and hybrid human-machine approaches) to discriminate the genuine reviews from the false ones. We find that overall a hybrid human-machine method works better than either human or machine-based methods for detecting fraud – provided the most salient features are chosen. Our process has implications for fraud detection across numerous domains, such as financial statements, insurance claims, and reporting clinical trials.
Complete list of metadata

Cited literature [39 references]  Display  Hide  Download

https://hal.inria.fr/hal-02510146
Contributor : Hal Ifip <>
Submitted on : Tuesday, March 17, 2020 - 2:56:14 PM
Last modification on : Tuesday, March 17, 2020 - 3:02:20 PM
Long-term archiving on: : Thursday, June 18, 2020 - 3:03:21 PM

File

 Restricted access
To satisfy the distribution rights of the publisher, the document is embargoed until : 2022-01-01

Please log in to resquest access to the document

Licence


Distributed under a Creative Commons Attribution 4.0 International License

Identifiers

Citation

Christopher Harris. Comparing Human Computation, Machine, and Hybrid Methods for Detecting Hotel Review Spam. 18th Conference on e-Business, e-Services and e-Society (I3E), Sep 2019, Trondheim, Norway. pp.75-86, ⟨10.1007/978-3-030-29374-1_7⟩. ⟨hal-02510146⟩

Share

Metrics

Record views

104