Maytal Messing - Researcher

Maytal has been working in eBay for a little over a year. These days, she is in the final stages of her M. Sc. in Computer Science in the Technion. You should know that online shopping is her expertise, even outside working hours. For Maytal, eBay is the ability to develop algorithms that impact hundreds of millions of people.

Hi Maytal, tell us about your job.
“I am part of a team that is responsible for developing a learning algorithm that can match any item a seller uploads, to the proper catalog product. The algorithm needs to analyze data that is composed of text and more structured data (like prices, specs etc.). The algorithm then needs to conclude when an item and a product are the same things. In order to do that, I use algorithms and advanced techniques of Deep Learning and Natural Language Processing (NLP)”.

This sounds complex. What are the challenges of your everyday work?
“One of the biggest challenges is to collect large quantities of samples of items and products that don’t match, and samples of those that do match, in order to train the algorithm to identify a match and a mismatch. Since human-based tagging is unrealistic what you need so many samples, we think of creative ways in order to create the samples automatically. For example, products have their own unique identifiers, so if we find a product and an item with the same identifier, it’s highly probable that it’s the same product and that they match. This assumption, along with other heuristics, we can use in mass, since we have huge amounts of data”.

What do you like best about your job?
“It’s nice to know that what we develop here, influences the shopping experience of tens and hundreds of millions of users. I really enjoy working on a subject that I find fascinating – which is also the subject of my thesis – quantifying semantic similarities between two entities. My thesis was about examining two general entities (people, places, etc.), using data from Wikipedia, and on eBay we are talking about products and items for sale – but there’s a lot of similarities in the challenges”.

Thank you, Maytal!