e5a532a8bc
- Introduced `test_nlp_extractor.py` with unit tests for the InstitutionExtractor, covering various extraction patterns (ISIL, Wikidata, VIAF, city names) and ensuring proper classification of institutions (museum, library, archive). - Added tests for extracted entities and result handling to validate the extraction process. - Created `test_partnership_rdf_integration.py` to validate the end-to-end process of extracting partnerships from a conversation and exporting them to RDF format. - Implemented tests for temporal properties in partnerships and ensured compliance with W3C Organization Ontology patterns. - Verified that extracted partnerships are correctly linked with PROV-O provenance metadata.
132 lines
4.3 KiB
Python
132 lines
4.3 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Test subagent-based NER extraction for heritage institutions.
|
|
|
|
This script tests using a coding subagent to perform Named Entity Recognition
|
|
instead of the pattern-based extractor.
|
|
"""
|
|
|
|
import sys
|
|
import json
|
|
from pathlib import Path
|
|
|
|
# Sample text with known institutions (same as V5 test)
|
|
SAMPLE_TEXT = """
|
|
# Dutch Heritage Institutions Discussion
|
|
|
|
## Valid Institutions (should be extracted)
|
|
|
|
The Van Abbemuseum in Eindhoven is a renowned modern art museum with an
|
|
ISIL code NL-EhdVAM. Founded in 1936, it holds one of the finest collections
|
|
of modern and contemporary art in Europe.
|
|
|
|
The Zeeuws Archief in Middelburg preserves historical records from the
|
|
province of Zeeland. ISIL: NL-MdlZA. It maintains extensive archival
|
|
collections dating back to the 13th century.
|
|
|
|
Historisch Centrum Overijssel (ISIL: NL-ZwHCO) in Zwolle serves as the
|
|
regional archive for the province of Overijssel.
|
|
|
|
## Organizations/Networks (should be filtered by V5)
|
|
|
|
The IFLA Library is an international organization that coordinates library
|
|
networks worldwide.
|
|
|
|
Archive Net is a global network connecting archives across multiple continents.
|
|
|
|
## Generic Descriptors (should be filtered by V5)
|
|
|
|
The Library FabLab is a makerspace facility open to the public.
|
|
|
|
University Library is part of the academic infrastructure.
|
|
|
|
## Geographic Errors (should be filtered by V5 country validation)
|
|
|
|
The National Museum of Malaysia in Kuala Lumpur houses extensive Southeast
|
|
Asian collections.
|
|
|
|
University Malaysia specializes in research and education.
|
|
"""
|
|
|
|
|
|
def main():
|
|
"""Test subagent-based extraction"""
|
|
print("=" * 70)
|
|
print("Subagent-Based NER Extraction Test")
|
|
print("=" * 70)
|
|
print()
|
|
|
|
print("Sample text:")
|
|
print("-" * 70)
|
|
print(SAMPLE_TEXT[:500] + "...")
|
|
print("-" * 70)
|
|
print()
|
|
|
|
print("This test demonstrates using a Task subagent for NER extraction")
|
|
print("instead of pattern-based regex extraction.")
|
|
print()
|
|
print("Expected output:")
|
|
print(" - Clean institution names (no mangling)")
|
|
print(" - Accurate type classification (MUSEUM, ARCHIVE, LIBRARY)")
|
|
print(" - Location extraction (city, country)")
|
|
print(" - ISIL code identification")
|
|
print()
|
|
|
|
# Instructions for using Task tool
|
|
print("=" * 70)
|
|
print("NEXT STEP: Use Task tool with subagent_type='general'")
|
|
print("=" * 70)
|
|
print()
|
|
print("Task prompt should be:")
|
|
print("-" * 70)
|
|
print("""
|
|
Extract ALL heritage institutions (museums, archives, libraries, galleries)
|
|
from the following text.
|
|
|
|
For EACH institution found, return JSON with:
|
|
{
|
|
"name": "Full institution name (no truncation)",
|
|
"institution_type": "MUSEUM | ARCHIVE | LIBRARY | GALLERY",
|
|
"city": "City name (if mentioned)",
|
|
"country": "2-letter ISO country code (if can be determined)",
|
|
"isil_code": "ISIL code (if mentioned, format: XX-XXXXX)",
|
|
"confidence": 0.0-1.0 (based on explicitness of mention)
|
|
}
|
|
|
|
Rules:
|
|
1. Preserve full names (e.g., "Van Abbemuseum", not "Abbemuseum")
|
|
2. Classify by primary function (museum, archive, library, gallery)
|
|
3. Extract city from context (e.g., "in Eindhoven" → city: "Eindhoven")
|
|
4. Determine country from city or explicit mentions
|
|
5. Find ISIL codes in patterns like "ISIL: XX-XXXXX" or "(ISIL: XX-XXXXX)"
|
|
6. Exclude:
|
|
- International organizations (IFLA, UNESCO, etc.)
|
|
- Networks/platforms ("Archive Net", "Museum Association")
|
|
- Generic descriptors ("University Library" without specific institution name)
|
|
- Academic departments (unless they have dedicated collections)
|
|
|
|
Return ONLY valid JSON array, no additional text.
|
|
|
|
TEXT TO ANALYZE:
|
|
""" + SAMPLE_TEXT)
|
|
print("-" * 70)
|
|
print()
|
|
print("Expected result: JSON array with 3 institutions")
|
|
print(" 1. Van Abbemuseum (Eindhoven, NL, MUSEUM)")
|
|
print(" 2. Zeeuws Archief (Middelburg, NL, ARCHIVE)")
|
|
print(" 3. Historisch Centrum Overijssel (Zwolle, NL, ARCHIVE)")
|
|
print()
|
|
print("Should filter out:")
|
|
print(" - IFLA Library (international organization)")
|
|
print(" - Archive Net (network)")
|
|
print(" - Library FabLab (generic descriptor)")
|
|
print(" - University Library (no specific name)")
|
|
print(" - National Museum of Malaysia (wrong country for Dutch focus)")
|
|
print(" - University Malaysia (wrong country)")
|
|
print()
|
|
|
|
return 0
|
|
|
|
|
|
if __name__ == "__main__":
|
|
sys.exit(main())
|