JDATA Absolute Semantic Recognition Challenge – Competition Summary and Winners
The JDATA Absolute Semantic Recognition Challenge concluded on July 15 with eight finalist teams presenting their models, the champion "Label Team" using BERT, judges highlighting data cleaning, model fusion, and the distinction between absolute and contextual words, and a detailed ranking of all participating groups.
The JDATA Absolute Semantic Recognition Challenge, organized by JD.com, ended on July 15 with eight finalist teams delivering live presentations in the C7 auditorium of JD Headquarters. The competition, which began on March 18, focused on identifying prohibited or non‑compliant language in product pages, distinguishing absolute terms that violate advertising regulations from contextual terms that are permissible.
Judges from JD Retail’s Platform Ecology Department and Technology & Data Middle Platform, including algorithm experts Zhou Mo, Liu Pengfei, Huang Mingxing, and Liu Tingting, evaluated the submissions. Zhou Mo praised the champion "Label Team" for its outstanding algorithm design, data cleaning, training‑set partitioning, metric optimization, and model fusion, noting that the team selected BERT, the current industry‑standard model.
Other notable teams were highlighted as well. Liu Pengfei commended the DOTA team for a complex, diverse, and robust solution adaptable to multiple scenarios, while Eric_Li was recognized for correcting OCR results and incorporating additional features such as unregistered title words, demonstrating strong business relevance and knowledge transfer.
The competition’s task—absolute semantic recognition—required participants to understand not only word meanings but also their contextual usage to reduce erroneous promotional language. For example, the word "most" in "the most suitable environment for white tea" is an absolute term violating advertising law, whereas "most genuine feelings" is a contextual term and permissible.
All competition data originated from JD’s real production environment, and the top‑performing models have the potential to be deployed in actual business settings. Award presenters, including JD Retail Platform Ecology’s Merchant R&D Head Fu Caibao and Data Application Platform Head Xie Xiang, emphasized the practical value of the algorithms for improving product page compliance and encouraging collaborative algorithm development.
Competition Rankings
Champion Team: Label Team – Members: Chen Xi, Ding Pan, Gao Junyuan, Li Han, Shi Xiaobin
Runner‑up Teams: cactus – Member: Chen Wei; Eric_Li – Member: Li Fei
Third‑place Teams: Fatty – Member: Xie Qunqun; DOTA – Members: Zhu Xiangyu, Liu Yuzhong, Wang Linlin; Chinabup – Member: Guo Xingyu
Outstanding Teams: Non‑Expert Parameter Tuning Squad – Members: Wang Xiaoyu, Yan Chuang, Yu Weiwe, Liu Na; GoID – Members: Liao Kexin, Yu Mingzhou
JD JDATA Zhihui Platform will continue to host more algorithm‑focused events in the future.
JD Retail Technology
Official platform of JD Retail Technology, delivering insightful R&D news and a deep look into the lives and work of technologists.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.