FatimahEmadEldin/MycoBase-Large-Scale-Text-to-SQL

MycoBase: A Biologically Literate Text-to-SQL Dataset

MycoBase is a synthetic but biologically accurate dataset designed for stress-testing Text-to-SQL systems. It represents a research information system for the study of fungi, covering everything from taxonomy and genomics to morphology and cultivation.

Dataset Highlights

• —Schema Complexity: 2,016 tables with over 9,000 foreign key relationships.
• —Data Volume: 320,270 rows of realistic mycology data.
• —Realistic Names: Includes tens of thousands of actual fungal species names, famous mycologists (authors), and scientific publication titles.
• —Format: Compatible with Spider and BIRD benchmark tools.

Repository Structure

• —database/mycobase/mycobase.sqlite: The SQLite database containing the full schema and data.
• —tables.json: Structural metadata (schema) in BIRD/Spider format.
• —evaluation_questions.json: Natural Language to SQL query pairs for training/evaluation (renamed from train.json).
• —mycobase_benchmark_summary.csv: A pipe-separated summary of questions, ground-truth SQL, and retrieval stats.
• —csv_exports/: Sample data for core tables (tax_species, specimens, etc.).
• —mycobase_sqlite_schema.sql: Full SQLite DDL script.

Example Questions

• —"Which species in our database contain amatoxins, and at what concentrations?"
• —"Show me ITS sequences with their species and the herbarium voucher they came from."
• —"Which Agaricales species are classified as deadly, and what is the mean amatoxin concentration reported for each?"

Dataset Generation

This dataset was generated using the MycoBase core framework, ensuring biological literacy across 57 fungal orders. Every table and column is modeled after real-world research workflows in mycology.

License

MIT