Yelp Review Filter
Yelp Review Filter
As many know, Yelp is a review website that was created to improve connections between people and businesses. The way Yelp works is that users can rate and submit feedback to businesses through the website. These reviews are publicly visible and serve the dual purpose of providing the business valuable feedback as well as giving prospecting customers an idea of the quality of services offered.
The significance of these kinds of interactions are high. For example, a terrible review for a restaurant can drive away many diners. On the other hand, many high reviews can cause the restaurant to appear at top of the search list and boost its business. As you can imagine, this automatically creates an environment for misbehavior. Restaurants may bribe people to give them high reviews. Restaurants owners might give themselves high reviews. Restaurants might even give other restaurants negative reviews. There are many scenarios that would create incentives for false reviews.
For this reason, Yelp has implemented many algorithms that can estimate the validity of reviews and decide whether to use them in the overall rating of a business. These algorithms are collectively known as the Yelp Review Filter. The Review Filter will analyze reviews and assign them some measure of validity. Valid reviews are more likely to show up at the top of a search for reviews, and less valid reviews are more likely to show up at the bottom or not even factor into a business’s rating at all. To understand this better, we can think of the interactions between customers and businesses as a pseudo hub-authority model, where the hubs are customers who endorse the authorities, which are the businesses.
The model I am about to detail will be simplified greatly and won’t represent the rating/review model perfectly. However, it can still serve as a good analogy. Let’s assume, for the sake of simplicity, that there are two types of edges that can be drawn between customers and businesses, an approval edge and a disapproval edge. This makes sense because customers can rate the business. Let’s also assume that we can draw edges between customers to represent social activity. Yelp accounts typically link to a social media account so this makes sense. An example of this graph would look something like this, where the black edges represent interactions between reviewers, red lines represent disapproval toward a business, and green lines represent approval toward a business.
The quality of a business can be represented by the inflow of green and red edges. Here, we can see that Goldie’s could be a good business because it has 3 green edges pointing into it. RPCC may be a lower quality business because it has 2 red edges pointing into it and only 1 green edge. We may also want to look into the legitimacy of the hubs that these red/green edges are coming from. The Yelp Review Filter’s job is to filter out any reviews it deems as suspicious and not genuine.
Duplicate reviews are usually a cause for concern. For example, if Alice were to write positive reviews for Goldie’s three times (represented by three green arrows pointing from Alice to Goldie’s), the Review Filter will find that suspicious and filter out some, if not all, reviews from Alice.
Another cause for suspicion might be that a user only gives 5 star reviews. Charles’s activity might raise some red flags because he has only given positive reviews (three green arrows coming out of Charles). This would be even more suspicious if there were, say 30 businesses, and Charles gave 5-star reviews to each one. Yelp’s Review Filter’s job is to filter out such reviews.
Finally, a user’s interactions with the community factor into the validity of his/her review. Reviewers with friends are far more likely to have credibility and their reviews will less likely be filtered. Charles has little interaction with the rest of the customers, so this may increase the chance that his reviews get filtered. A and B are connected so their reviews may hold a little more validity.
Sources:
http://www.reviewtrackers.com/learn-triggers-yelp-review-filter/
https://www.entrepreneur.com/article/235271