Detecting Online Review Fraud Using Sentiment Analysis

Start Date

15-4-2021 3:45 PM

End Date

15-4-2021 4:00 PM

Student's Major

Computer Information Science

Student's College

Science, Engineering and Technology

Mentor's Name

Rajeev Bukralia

Mentor's Department

Computer Information Science

Mentor's College

Science, Engineering and Technology

Description

With the exponential increase of e-commerce markets, reviews on products have become a substantial advocate for a shop and product’s reputation. Consequently, fake reviews have become a way to ploy customers into trusting the credibility of a product. On account of this, fake reviews have been a topic of research for as long as e-commerce stores have had review sections. Nevertheless, there still has not been an efficient solution found to detecting these fake reviews. With this research we hope to gain insights to continue the development of detection techniques. To do this we have explored the accuracy of sentiment analysis on book review data through quantitative research. To complete this analysis, we have found the polarity score of each of the reviews and correlated it to the star rating of the review. Results from this research have found that the polarity score of the review is not an effective value to use when detecting fake reviews. This is due to the disconnect of the language used in the review to the rating given. This disconnect could be a cause of several factors, such as the limitations in ontology used in sentiment analysis and the review language not reflecting the overall sentiment of the rating given.

This document is currently not available here.

Share

COinS
 
Apr 15th, 3:45 PM Apr 15th, 4:00 PM

Detecting Online Review Fraud Using Sentiment Analysis

With the exponential increase of e-commerce markets, reviews on products have become a substantial advocate for a shop and product’s reputation. Consequently, fake reviews have become a way to ploy customers into trusting the credibility of a product. On account of this, fake reviews have been a topic of research for as long as e-commerce stores have had review sections. Nevertheless, there still has not been an efficient solution found to detecting these fake reviews. With this research we hope to gain insights to continue the development of detection techniques. To do this we have explored the accuracy of sentiment analysis on book review data through quantitative research. To complete this analysis, we have found the polarity score of each of the reviews and correlated it to the star rating of the review. Results from this research have found that the polarity score of the review is not an effective value to use when detecting fake reviews. This is due to the disconnect of the language used in the review to the rating given. This disconnect could be a cause of several factors, such as the limitations in ontology used in sentiment analysis and the review language not reflecting the overall sentiment of the rating given.