Agentic data science with the Data Agent Kit
cloud.google.com·7h ago
AI Research
arXiv CS.AI·Aug 13·5 min read
BEST-KAG: Enhancing Question Answering of Building Engineering Standards with Multimodal Knowledge Graph Modeling and Large Language Model
TL;DR
BEST-KAG is a novel framework that enhances question answering for building engineering standards by integrating a multimodal knowledge graph and a large language model. It significantly outperforms existing models, achieving up to 74.01% improvement in evaluation metrics.
✦ Why It Matters
Implement BEST-KAG to improve the accuracy and traceability of question answering in building engineering standards today.
Key Takeaways
- Adopt BEST-KAG for enhanced question answering in engineering standards.
- Utilize the multimodal knowledge graph for better data organization.
- Leverage the hybrid rule-LLM approach for scalable knowledge extraction.
- Focus on clause-grounded responses for improved traceability.
How It Works
BEST-KAG integrates a multimodal knowledge graph (MKG) that represents the hierarchy and relationships of building engineering standards. It combines rule-based methods with large language models (LLMs) to extract knowledge, creating a vast graph that supports complex question answering.
The architecture allows for retrieval of specific clauses, ensuring that answers are not only accurate but also traceable back to the original standards.