E-Commerce Product Categorization with Watson

Zia Mohammad
IBM watsonx Assistant
3 min readApr 17, 2018
Image Source: Carl Raw

The founding of Amazon, E-Bay, and other companies allowed for the rise of e-commerce. This new type of industry would cement itself as one of the dominant forces in retail. New ways to buy and sell became standard and customers could find anything online as their searches were optimized with technologies such as IBM Watson.

How can we list our products online in such a way that people can find them?

Departments > Electronics > Headphones. While looking online for headphones, I found these terms floating through my head. I began to ask myself: How are e-commerce sites or websites able to sort and organize millions of items?

Photo by Pro Church Media on Unsplash

All those reviews, product descriptions, ratings, must have some level of technical complexity. Given the volume of items, it can’t be as simple as cutting out labels and pasting them on a product.

What methods exist for retailers to easily categorize product descriptions?

Eureka!

While talking with fellow IBMer and cognitive consultant, Eric Hensley, we had our Eureka moment. Although we didn’t find ourselves parading the office shouting Eureka, we did find a solution that leveraged Watson’s Natural Language Classifier to solve this enterprise problem.

Eric’s sample app helped to address the issue of product categorization. By leveraging the Watson Natural Language Classifier service, he was able to train the application to categorize online product descriptions. A similar application could be integrated in a companies processes to automatically categorize items by product descriptions, simplifying the backend processes.

Additionally, a company building upon this simple web app can include product reviews, ratings, and other forms of data helping internally and externally. Tags can be automatically be applied to various product descriptions or reviews which will help customers find their ideal product.

Product Categorization Sample App

Bookshelf classification example

Try out the sample app or build your own: Product Categorization Demo

Before you go, here are some quick tips:

  • You can input your own product description. At the bottom, you’ll see multi-level classification on your product description
  • Can’t think of a product description? Find something on Kohl’s.com and watch Watson Natural Language Classifier classify it. Try editing the input!
  • Machine learning allows the classifier is able to handle product descriptions not seen in the original training data. Give it a try!

Building a version of this for your enterprise? Fork the Github repo or watch the tutorial video below:

Video Tutorial: Make your own e-commerce product classifier

--

--

Zia Mohammad
IBM watsonx Assistant

Senior Product Manager @ AWS Quantum | Futurist | Global Citizen | Adventurer | Buckeye