Harish Karnick

Keynote at ADCOM 2016

Title : Natural Language Processing in E-Commerce: A Case Study

Abstract:  E-commerce services like Amazon, Flipkart, Snapdeal and many others are evolving into market places where independent buyers and sellers meet and the the e-commerce company essentially plays the role of an intermediary regulator and guarantor. Some companies like eBay have been like this from the start. The number of products on offer on such platforms ranges in the multi-millions. Unfortunately, the active set of products that typical users end up actually viewing is usually in the several thousands. So, no one even sees the vast majority of products.

One way to mitigate this viewing problem is to change the way in which a user searches for and reaches a product. The intuition is to do to products what has been done to library books – create a hierarchical catalogue. Most of the time one can easily navigate the catalogue in the library and find a book even when there are millions of books. The hope is that if products are catalogued suitably then they will be equally easy to find and therefore help e-commerce platforms to do better. The task is to define a hierarchical tree of category labels (often called an ontology) and tag a product with a leaf label. This automatically tags the product with those labels that are in the path from the leaf to the root.

The goal of a project we have been associated with recently is to take textual descriptions of products and/or titles created by the seller and predict a path as a classification label for the product. I will discuss the difficulties we faced in this project and how we have overcome some of them to get reasonable results. This is still work in progress.


Harish Karnick teaches in the Department of Computer Science and Engineering at IIT Kanpur. His interests are machine learning, automated & commonsense reasoning, cognition and more generally artificial intelligence. Programming languages, their history and evolution is a hobby and LISP is a particular favourite. He is an alumnus of IIT Kanpur.