“Category prediction system” 

Automatic classification tool for predicting the correct category for a given product title and description. Under the hood is a combination of two different neural models – deep belief networks and deep autoencoders.

High scalable to categorize a large e-commerce dataset consisting of 200 millions of products and a 5-level taxonomy tree containing 28K categories.

Had a relative speech at NVidia’s GTC 2017:

Large-scale product categorization with deep models in Rakuten

Posted in Project

Leave a Reply