Idea
Multimodal classification platform for e-commerce businesses to improve product categorization and discover fine-grained categories efficiently.
Research Paper
Core Innovation
This paper presents a multimodal hierarchical classification framework combining text, image, and joint vision-language features to overcome platform heterogeneity and taxonomy limitations. It leverages CLIP embeddings for high accuracy and introduces a self-supervised recategorization pipeline to identify fine-grained categories. The two-stage inference pipeline balances accuracy and computational cost, enabling industrial scalability.
Market Size (TAM)
$10–20B TAM, $2–5B SAM; assumption: global e-commerce market growth and increasing demand for automated product categorization.
Potential Customers & Pain Points
- E-Commerce Platforms Struggling With Inconsistent Product Taxonomies
- Fashion Retailers Needing Accurate Cross-Platform Categorization
- Marketplaces Seeking Scalable Categorization Solutions
Business Model
SaaS platform offering API access for product categorization and recategorization with tiered pricing based on volume and features.
Competitive Landscape
- Amazon Product Categorization
- Google Cloud Vision API
- Clarifai
Implementation Challenges
- Integration complexity across diverse platforms
- Data privacy and proprietary taxonomy concerns
- Computational cost for large-scale deployment
Validation Strategy
- Pilot integration with mid-size fashion e-commerce platforms
- Measure categorization accuracy and cost savings versus existing methods
- Iterate model based on user feedback and scalability tests
Research Paper Overview
Cross-Platform E-Commerce Product Categorization and Recategorization: A Multimodal Hierarchical Classification Approach
Summary
This study develops a multimodal hierarchical classification framework integrating textual, visual, and joint vision-language features to address platform heterogeneity and taxonomy limitations in e-commerce product categorization. Using 271,700 products from 40 fashion platforms, it achieves high accuracy with CLIP embeddings and introduces a self-supervised recategorization pipeline to discover fine-grained categories. The framework balances accuracy and cost via a two-stage inference pipeline and demonstrates industrial scalability in EURWEB's platform.