arxiv: https://arxiv.org/pdf/2412.01837
LLM-PKG, a novel framework for generating explainable product recommendations in e-commerce. It leverages large language models (LLMs) to create a product knowledge graph (PKG), which is then used to provide recommendations along with clear rationales. The framework incorporates prompt engineering techniques to ensure high-quality LLM outputs and employs rigorous validation and refinement methods to mitigate LLM hallucinations. A/B testing demonstrates that LLM-PKG significantly improves user engagement and sales compared to traditional methods. The system efficiently maps LLM outputs to real-world products and supports both item-centric and user-centric recommendation scenarios.
Created with NotebookLM
LLM-PKG, a novel framework for generating explainable product recommendations in e-commerce. It leverages large language models (LLMs) to create a product knowledge graph (PKG), which is then used to provide recommendations along with clear rationales. The framework incorporates prompt engineering techniques to ensure high-quality LLM outputs and employs rigorous validation and refinement methods to mitigate LLM hallucinations. A/B testing demonstrates that LLM-PKG significantly improves user engagement and sales compared to traditional methods. The system efficiently maps LLM outputs to real-world products and supports both item-centric and user-centric recommendation scenarios.
Created with NotebookLM
- Catégories
- E commerce Ebay
Commentaires