Ido Guy, Head of Research dep. in RecSys conference.
We are proud to present Ido Guy, eBay’s Research department manager. Ido has been an active participant in the steering committee in Recsys* of the ACM for the past six years, and this year he represented us in a presentation in front of around 800 researches from all over the world.
Ido, we are curious – what was the presentation about?
“This year, I talked about how we connect eBay’s buyers and sellers, so they can ‘speak’ the same language. I focused on the challenges and they way we deal with them in eBay’s research department”.
Tell us about the challenges.
“One challenge is the Matching issue – the ability to exactly match between items our sellers upload to the site, and the products on eBay’s catalog – which is huge. Another challenge is Duplication – there are many times in which sellers upload items with a title that doesn’t match another item, which is identical. It’s the same item, but not the same title. There’s aldo Aspect Extraction – how much can we understand from the item’s description, written by the sellers. What part of it is the name of the item (i.e., the title), which category does it belong to, what’s the brand name, color, etc. There are many more fascinating challenges, but we’ll save them for the next post 🙂 “.
You made us even more curious! How did you guys in the research department overcome these challenges?
“We combine different methods of language processing and computer vision, and apply them on different data supplied by the sellers. Some of these techniques include machine learning for ranking; identifying an object out of an image; identifying an entity from text; automatic translation and more. These techniques enables us to create a better representation of the various users and items in the site, and calculate the similarities between them more effectively”.
Ido, thanks for representing us and good luck!
* Every year, RecSys brings together hundreds of researchers from the industry and the academic world. The researchers share information about the hottest researches on Recommender Systems.