Are Online Consumer Reviews Credible? A Predictive Model based on Deep Learning

Ehsan Abedin, Antonette Mendoza, Shanika Karunasekera

Research output: Contribution to conferencePaperpeer-review

Abstract

As the importance of online consumer reviews has grown, the concerns about their credibility being damaged by the presence of fake reviews have also grown. Extant literature reveals the importance of online reviews for consumers. Yet, there is a lack of research in the literature that considers consumer perception while developing a predictive model for the credibility of online reviews. This research aims to fill this gap by combining two different streams in the literature namely human-driven and data-driven approaches. To do so, we use two datasets with different labelling approaches to develop a predictive model, the first one is labelled based on the Yelp filtering algorithm and the second one is labelled based on the crowd’s perception towards credibility. Results from our predictive model reveal that it can predict credibility with a performance of 82% AUC, using reviews’ attributes namely, length, subjectivity, readability, extremity, external and internal consistency.

Original languageEnglish
Publication statusPublished - 2022
Externally publishedYes
Event33rd Australasian Conference on Information Systems: The Changing Face of IS, ACIS 2022 - Melbourne, Australia
Duration: 4 Dec 20227 Dec 2022

Conference

Conference33rd Australasian Conference on Information Systems: The Changing Face of IS, ACIS 2022
Country/TerritoryAustralia
CityMelbourne
Period4/12/227/12/22

Keywords

  • Deception Detection
  • Fake Reviews
  • Online Consumer Reviews
  • Review Credibility

Fingerprint

Dive into the research topics of 'Are Online Consumer Reviews Credible? A Predictive Model based on Deep Learning'. Together they form a unique fingerprint.

Cite this