eBay Sets German Auto Part Identification Challenge For 2025 University Machine Learning Competition
eBay's annual university competition tasks US college students with solving auto part identification challenges specifically for the company's German marketplace, with winners vying for a spot in the 2026 eBay summer internship program.
Attention all students! We're thrilled to announce the return of eBay's 7th Annual University Machine Learning Competition, and this year, we're diving into the exciting world of e-commerce datasets once again.
A huge thank you to everyone who participated in the 2024 competition—the competitors were all amazing! We can't wait to see what kind of buzz this year's challenge will create.
Here's the exciting part: Winners will have the opportunity to secure a summer internship with eBay for 2026! That's right, we're always on the hunt for top talent, and our internship program could be your stepping stone to a full-time role with us.
You've spoken, and eBay has listened! Following the fantastic feedback from previous participants, we're diving back into the Motors scene this year.
Here's your mission: Dive deep into the realm of motor vehicle parts and accessories. Your task? Develop a model capable of accurately identifying and labeling named entities in eBay item titles. Named entities are those key words or phrases that refer to people, brands, organizations, locations, styles, materials, patterns, product names, sizes, and more.
For car parts, named entities might be things like make and model of the car that the part fits, manufacturer of the part, type of part, what’s included in a kit, and more.
Named Entity Recognition (NER) is the machine learning technique that automatically identifies and labels these important entities in a text. In the e-commerce world, NER is vital for processing product titles and descriptions, customer queries, reviews, and anywhere else important data needs to be extracted from raw text.
At eBay, we use NER in numerous ways, especially for extracting key details from listings (for sellers) and search queries (for buyers). NER is essential for transforming unstructured text into structured data. This challenge focuses on extracting information from listings.
To showcase eBay's global reach, this year's challenge data will be sourced from a non-English site—specifically, listings from eBay's German site. In Germany, eBay is the No.1 marketplace for vehicle parts and accessories.
Think you've got what it takes to excel in the NER arena? Show us your skills, and let's make this competition unforgettable!
Students who enter this year's competition will be asked to extract and label aspects like item brand, make and model of the car that the part fits, manufacturer of the part, type of part, what’s included in a kit, and more using a dataset of German language listing titles for items on eBay.de.
eBay provided an example in English to illustrate the concept using different types of items.
The actual data set for the competition will consist of 2 million randomly selected unlabeled item titles from eBay Germany, all of which are related to the "Car Brake Component Kits" and "Car Engine Timing Kits" categories.
Interestingly, eBay has put an AI disclaimer on the challenge this year, advising participants they are not allowed to enter any eBay data into third party hosted AI platforms or any tools which limit commercial use of AI generated code.
Use of AI
eBay data MUST NOT be inputted into third party hosted platforms (including Generative AI tools, such as ChatGPT, Bing Chat, etc.) where the data is subject to retention and use by the third party platform.Furthermore, if AI tools are used, the resulting code, data, or other generated content MUST NOT limit commercial use of the code, data, or any model containing or depending on the code.
In particular, the winning method MUST NOT require any third-party licenses to use the data, model, or software, nor depend on anything else that would prevent eBay from using the winning method for commercial or any other purposes.
However, open source software and public domain algorithms and data MAY be used in the winning method if licensed for commercial use and/or publicly available under a permissive Open Source Initiative approved license (MIT, BSD, Apache 2.0, etc.).
Sign up is open from May 1st - September 10th, with a deadline for submissions of November 3rd and winners will be chosen and notified on November 14th.
Details:
