Yelp provides a convenient way to discover restaurants with top reviews by users. However, besides these reviews, most customers are also interested in information about the businesses themselves, such as menus for restaurants. Customers also often require information about a restaurant’s most popular dishes in order to make a decision on what to order, especially if it’s their first time at the establishment.

We propose a system to automatically detect and extract dishes from reviews and create a list of popular dishes for each restaurant. Our solution aims to be independent of the availability of menus, which are often not available or difficult to source. The only requirement would be a large enough corpus of reviews to allow our system to accurately extract the top dishes information. The problem can be broken down into named entity recognition, where we treat food as a category of entity. This system would be relevant to all types of restaurant goers, from the casual diners who are new to the city or are exploring the area for the first time, to avid food connoisseurs trying out the food at a well rated restaurant for the first time. This would also be valuable to restaurant owners to keep track of the most sold dishes, without having to sift through thousands of reviews.