Problem with Yelp: * Lots of irrelevant material when it comes to deciding what to order Solution: * DOOR - Deep Object-Oriented Recognition * A quick way to process information and decide what to order and what to avoid Method: * Long Short-term Memory Embedded Neural Network that really understands the concept of food * Essentially insert text (one sentence or 10 million reviews) and we will extract all of the food words
* The network doesn't use explicit dictionary (or else we won't be able to discover food terms not in list)
* Network learns to categorize each food based on context, term freuquency, linguistic relationship to
search concept and neighboring words. The actual word is the least important aspect)
* In many ways this is how we learn about concepts. If I gave you a sentence with a foreign foodand told you
to find the word, most of you will be able to do. The algorithm works the same way.
Built With
- amazon-web-services
- azure
- beaker
- ionic
- javascript
- python
- theano
Log in or sign up for Devpost to join the conversation.