Sponsored Link
Email This Post Email This Post

eBay Uses Machine Learning to Classify Listings Based on Their Titles

eBay Uses Machine Learning to Classify Listings Based on Their Titles

eBay once again turned to university students to help it solve real-world problems. This week, the company announced the winners of its 4th Annual Machine Learning Challenge.

Last year, eBay tasked students with helping it improve the delivery estimates it displays to show shoppers approximately how long it will take for them to receive their items.

This year, it tasked students with helping it analyze eBay listing titles in order to help it classify listings. eBay explained:

“This year, applicants were given the challenge of building a model that can accurately extract and label the named entities in the dataset of item titles on eBay. Those “named entities” might include brands, locations, styles, product names, colors, materials, sizes, and other semantic strings, words and phrases that can help classify an item.”

eBay said the students used a machine-learning process called Named Entity Recognition (NER) and said applicants tackled the very real-world challenge any ecommerce platform faces: “how do we extract structured data from unstructured sources like listings?”

eBay announced two winners out of the 591 teams (consisting of 900 students from 157 colleges) that entered this year: Rupashi Sangal and Sanjayan Pradeep Kumar Sreekala, students at the University of California, San Diego

One of the attractions in entering the contest, according to Sangal, was the opportunity to work with a real-world dataset. eBay provided the students with a data set of eBay women’s handbag listings.

Sangal and Sreekala have accepted internships with eBay’s Structured Data Applied Research team in San Jose for the summer of 2023.

More information about this year’s eBay Machine Learning Challenge for University Students is available on the eBay corporate blog.

Ina Steiner on EmailIna Steiner on LinkedinIna Steiner on Twitter
Ina Steiner
Ina Steiner
Ina Steiner is co-founder and Editor of EcommerceBytes and has been reporting on ecommerce since 1999. She's a widely cited authority on marketplace selling and is author of "Turn eBay Data Into Dollars" (McGraw-Hill 2006). Her blog was featured in the book, "Blogging Heroes" (Wiley 2008). She is a member of the Online News Association (Sep 2005 - present) and Investigative Reporters and Editors (Mar 2006 - present). Follow her on Twitter at @ecommercebytes and send news tips to ina@ecommercebytes.com. See disclosure at EcommerceBytes.com/disclosure/.

9 thoughts on “eBay Uses Machine Learning to Classify Listings Based on Their Titles”

  1. Cheap workforce? Explains causes of sales disruption and login issues. My listings are not showing up in searches. When i search for comps I can only see 5 items sold which may make me price items inadequately or incfluence decision of not putting item for sale thinking it is not selling or selling too low. Go figure.

  2. I use structured data to create listing titles. Unfortunately, some titles are too long and must be shortened or reordered for readability. I doubt if an algorithm will be of much use unless the category and Item specifics are the input to the algorithm. Structured data mirroring the title and description is posted in item specifics. It is unfortunate that some eBay required Item Specifics in the Stamps category are corrupted abbreviations instead of English phrases. It is also unfortunate that eBay does not use the Scott catalog number appearing in most US listings as an Item Specific. Catalog number identify defining properties like design, denomination, color, paper and manufacturing characteristics. Item Condition, grade, and quality Item Specifics and the Condition Description help determine relative pricing for each individual stamp.

    Perhaps the students would be better served to study thermodynamics. Starting with structured data to produce listings would provide better results than starting with unstructured data. How often does the local paint store offer to unmix paint?

  3. Thank you for GSM⁉️
    “Government Sponsored Money IE:(Earn as you learn programs ‼️)
    So sleaze -bay Executives can sit @walkers on government & sellersmoney ‼️

  4. Yes. Yes. Fixing glitches promptly and creating board to report them that reflects also that the reported issues are being worked on.
    Fixing issue with staying logged in would be good start. „Stay signed in” preferences are being ignored you are logged out out every couple of minutes and you have to complete 2 captcha in raw of picking tall buildings.

  5. Why not develop an AI BOT to write titles and listings from catalog item specifics and the condition description? Could a more advanced version infer from listing images? The eBay website is going in the wrong direction with the “Advent Calendar” approach “optimized” for cell phones. Who can find errors and get an overview when so much is hidden?

    Can listing BOTS from third-party applications supplant the eBay supplied game of battleship that masquerades as a selling interface?

    How would eBay AI to deconstruct listings interpret AI created listings?

Leave a Reply