How I achieved 83% accuracy in auto-labelling on production

By Jiazhen Xie | 4 January 2018

Tags: machine-learning, automation, ai, experience

Description Previously in Dressipi, the Auto-Labelling worked as extracting feature keywords from garment title and description, then passed onto stylists to moderate the auto-labelled features. That means engineers have to maintain a list of keywords for each partner on per category basis, then requires human resources to inspect all of them. We would like to predict the correct garment features from the garment title and description, by using Machine Learning. Furthermore, one of the targets i...

← Back to Home