Design, Evaluation, and Production of a VOC Tagging System for Taobao User Experience
Taobao’s Technical Industry Data team designed a four‑level VOC tagging hierarchy to unify fragmented user‑feedback sources, evaluated label similarity with vector‑based distance matrices, optimized tag groups via entropy‑driven re‑grouping, built a stacking ensemble of FastText and TextCNN achieving over 90% accuracy, and deployed an automated production pipeline that generates tags, maintains ODPS tables, and provides APIs for rapid experimentation.