Research Paper

Core Innovation

This paper introduces a fully automated AI agent-driven framework leveraging large language models to create and refine ontologies and populate product knowledge graphs directly from unstructured descriptions. Unlike prior manual or rule-based methods, it requires no predefined schemas and ensures semantic coherence and scalability.

Why It Matters

E-commerce platforms face challenges managing vast unstructured product data, limiting search, recommendations, and analytics. Automating knowledge graph construction improves data organization and accessibility, reducing manual effort and enabling scalable, intelligent product data use. This transformation supports better customer experiences and operational efficiency across retail.

Potential Customers & Pain Points

• E-commerce platforms – Difficulty structuring diverse product data
• Retail analytics providers – Need high-quality scalable product knowledge
• Recommendation system developers – Require comprehensive product attributes
• Product information managers – Manual ontology and data curation overhead.

Business Model

SaaS subscription offering API access and platform tools for automated product knowledge graph construction, with tiered pricing based on data volume and customization needs.

Competitive Landscape

• Diffbot
• Stardog
• PoolParty
• Ontotext

Implementation Challenges

• Integration complexity with diverse e-commerce platforms
• Ensuring accuracy and consistency across heterogeneous product categories
• Adoption resistance due to existing manual workflows

Validation Strategy

• Pilot deployment with mid-to-large e-commerce retailers to measure integration ease and data quality improvements
• Benchmarking against manual and rule-based KG construction methods on diverse product categories
• User feedback collection from product managers and data scientists on usability and impact