Reproduce “Text-to-SQL for Enterprise Data Analytics” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Albert Chen, Manas Bundele, Gaurav Ahlawat, Patrick Stetz, Zhitao Wang, Qiang Fei, Donghoon Jung, Audrey Chu and 10 more
LinkedIn developed an enterprise-scale Text-to-SQL solution that integrates a comprehensive knowledge graph and a multi-agent system to enable non-SQL experts to self-serve data analytics from a vast, dynamic data lake. The system maintains over 300 weekly active users, with 33% of chat sessions resulting in code being pasted into the SQL editor.
Key insights
- Deep semantic understanding, captured by a rich Knowledge Graph beyond just schema, is crucial for generating semantically correct queries in complex enterprise data environments.
- Robust error correction mechanisms, such as an iterative query validation loop and a specialized Researcher LLM Agent for hallucination, are essential for achieving high query correctness and avoiding invalid outputs in real-world scenarios.
- A multi-agent architecture and a comprehensive user interface, extending beyond pure query generation to include data discovery, explanation, and debugging, significantly enhance user utility and adoption in enterprise settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.