This project won 1st place at HackRice 9! (presentation and demo link)
Have you ever ate out at a new spot and been unsure what to order? Don't want to have to turn the waiter down to ask for more time to look through countless yelp reviews all recommending different things? Or are you at a spot you've been to before? We wanted something that would take our custom preferences, past reviews, and perception of past meals we'd eaten and find us an optimal order.
What it does
Me.nu is a web application that recommends users what to eat when dining out. It takes into account reviews on other online platforms and user-specific preferences and past experiences to create the perfect custom order at any restaurant.
- Uses Google Vision API's Optical Character Recognition to parse a menu
- Filters results based on user preferences (budget, dietary. restrictions, etc)
- Collects reviews from multiple services, including Google Reviews and Yelp, to create an accurate prediction of the best dishes at a resteraunt.
- Factors in a user's previous Yelp reviews to help predict which menu items would be most appealing to them
- Creates a personalized account tied to a user's Yelp account to help learn a user's preferenences.
- Pulls food items from online libraries and dictionaries to give a full featured prediction of a user's tastes.
- Uses simplified relational calculus to calculate a score for each individual menu item.
How I built it
OCR with Google Cloud Vision, Extracting Google reviews with Google Maps Places API, Beautiful Soup and Xpath to extract Yelp reviews. Deployed with Heroku.
Challenges I ran into
Getting our web app to deploy with Heroku, optimizing OCR to provide accurate results.
Accomplishments that I'm proud of
Building a fully functional full stack web app that is fully functional. Learning new tools, and making an app that we think people will really enjoy!
What's next for Me.nu
Incorporating more advanced sentiment analysis for the reviews to more accurately rate dishes. Improving the UI.