NEWS

Machine Learning

Hone your Data Science Skills and Compete in the Rakuten 2020 Data Challenge

2020.07.11

The Rakuten Multi-modal Product Data Classification and Retrieval challenge, organized by RIT, is still ongoing. This year’s challenge is part of 2020 SIGIR Workshop on e-Commerce, In conjunction with ACM SIGIR Conference on Research and Development in Information Retrieval. Rakuten France has released a sampling of approximately 100K product titles from their entire catalog. The goal is to predict each product’s type code as defined in the catalog of Rakuten France and to retrieve the images corresponding to the products. We extended the Stage 1 – Model Building phase to July 15. Under this first stage, participants build and test models using Rakuten training data sets. At the time of writing, there have been new submissions for both product-type code classification and for cross-modal retrieval. The highest score remains a 91.83, using macro-F1 as evaluation.  The highest score is 50.23 under task 2, retrieval, evaluated on recall at 1 (R@1).

Additionally, we are accepting system description papers until July 17. Papers should address implementation details, including data preprocessing, error analysis, model descriptions, and any additional data used from eternal sources. Papers should be 2-4 pages in length and must be formatted according to the ACM SIG proceedings template

With a week to go, there is still opportunity to compete. The challenge presents interesting research aspects, including the cataloging of product labels and images in e-commerce marketplaces and deploying multimodal approaches. It gives researchers the chance to showcase their machine learning skills, build their portfolio (for both experienced and new practitioners), and continue to establish themselves in the field. Participants can also see how they fare in comparison.  If this sounds like you, and you’re fascinated with working with a huge repository and conquering analysis, consider joining us. You can register here for participation and submit papers here.

Finally, if you have any additional inquiries, please write Hesam Amoualian (hesam.amoualian@rakuten.com) or Parantapa Goswami (parantapa.goswami@rakuten.com) of RIT Paris, organizers of the Rakuten Data Challenge.

Copied! instagram