PIPER: Content-Based Table Search via profiling and LLM-Generated Pseudoqueries

📰 ArXiv cs.AI

Learn how PIPER uses LLM-generated pseudoqueries for content-based table search, enabling more effective dataset reuse and analysis

advanced Published 19 May 2026
Action Steps
  1. Build a table profiling system using LLMs to generate rich representations of tables
  2. Generate pseudoqueries using LLMs to capture table content and context
  3. Configure a search system to utilize pseudoqueries for content-based table search
  4. Test the search system using a dataset of tables with varying schema and cell values
  5. Apply PIPER to a real-world dataset to evaluate its effectiveness in facilitating dataset reuse and analysis
Who Needs to Know This

Data scientists and researchers working with large tabular datasets can benefit from PIPER's content-based search capabilities, improving dataset discovery and reuse

Key Insight

💡 LLM-generated pseudoqueries can effectively capture table content and context, enabling more accurate content-based table search

Share This
🚀 PIPER: Content-Based Table Search via LLM-Generated Pseudoqueries 🚀

Key Takeaways

Learn how PIPER uses LLM-generated pseudoqueries for content-based table search, enabling more effective dataset reuse and analysis

Full Article

Title: PIPER: Content-Based Table Search via profiling and LLM-Generated Pseudoqueries

Abstract:
arXiv:2605.18199v1 Announce Type: cross Abstract: The rapid growth of tabular datasets in data lakes, data spaces, and open data portals makes effective dataset search essential for reuse and analysis. Existing search systems rely mainly on metadata, which is often incomplete or low quality, especially for tables whose meaning depends on both schema and cell values. Recent advances in Large Language Models (LLMs) enable richer, content-based representations of tables. However, prior LLM-based re
Read full paper → ← Back to Reads