Discrepancy detection between actual user reviews and numeric ratings of Google App store using deep learning

Nowadays online reviews play a significant role in influencing the decision of consumers. Consumers show their experience and information about product quality in their reviews. Product Reviews from Amazon to Restaurant Reviews from Yelp are facing problems with fake reviews and fake numeric ratings. Online reviews typically consist of qualitative (text format) and quantitative (rating) formats. In the case of Google Play store fake numeric ratings can play a big role in the success of apps. People tend to believe that a high-star rating may be significantly attached with a good review. However, user star level rating information does not usually match with text format of review. Despite many efforts to resolve this issue, Apple App Store and Google Play Store are still facing this problem. This study proposes a novel Google App numeric reviews & ratings contradiction prediction framework using Deep Learning approaches. The framework consists of two phases. In the first phase, the polarity of reviews are predicted using sentiment analysis tool to build ground truth. In the second phase, star ratings are predicted from text format of reviews after training deep learning models on ground truth obtained in the first phase. Experimental results demonstrate that based on actual user reviews the proposed framework significantly predicts unbiased star rating of app.

Automated summary

Online reviews are influential but plagued by fake content. This study introduces a novel framework using deep learning to predict unbiased star ratings of apps based on user reviews. The framework involves two phases including sentiment analysis and text-based star rating prediction, showing significant accuracy in predicting unbiased app ratings.