Competitive Product Matching Based on NLP

M.Sc. Dissertation, performed in NTU & Lazada

Features:

  • Proposed a two stage text-based product matching framework to retrieve competitive product groups on E-commerce platform
  • Inspired by InfoNCE, designed a novel loss function PNB Loss for fine-tuning Sentence-BERT in Ecommerce domain, enabling effective text representation learning on product titles with weakly-supervised labels