Review2

Motivations

When it comes to finding out more about new businesses and locations, online reviews are one of the first things people tend to look at to assess its reputation. However, not all reviews are equal in quality, and some are more useful than others.

The impact of the overall quality of reviews is threefold: For users, high quality reviews allow them to make good decisions For businesses, relevant and truthful reviews allow fair representation. Malicious reviews can unfairly damage their reputation For platforms, bad reviews diminishes the credibility of their platform, affecting user retention

We've developed an ML-based solution for Track 1: Filtering the Noise: ML for Trustworthy Location Reviews which evaluates the quality and relevancy of Google location reviews.

What is Review2

Review2 is a chain of ML-based models which sequentially evaluates a given review to detect for quality related policy violations and flag it out for further review if needed.

Our solution runs in two stages: In the first stage, we use simple models and rules based filters to flag out our more simple policies such as spam, ads, and hate content. In the second stage, we use some more complex models to filter for some more indistinct factors such as how relevant the review is and if it was written second-hand.