Using Public Taxonomies

A Yuletide present

Dec 24, 2025

As I was working on a recent SHACL/AI taxonomy post, one of the major problems I kept encountering was how to represent named entities. Named Entity Recognition (NER) has always been more an art than a science, in significant part because there is no algorithmic mechanism available that will allow you to differentiate a name of a company from a name of a person or thing definitively.

The idea of using public taxonomies for NER and classification has, of course, been around for several decades. Still, it hasn’t been widely adopted in large part because there hasn’t been sufficient use case to justify the maintenance headaches of such taxonomies. However, I think this is changing. As more and more organisations need to be able to specify the SAME things using the same identifiers, there is a growing awareness of (and investment in) public IRI knowledge bases.

Note, again, taxonomies by themselves are not necessarily ontologies: an ontology is generally an agreed-upon standard for the structure of data, something that’s very useful, of course, but something that doesn’t necessarily tell you anything about identity. A taxonomy (especially a public taxonomy), on the other hand, is all about identity - how do you provide a common name for an entity that is nonetheless universal and easily referenceable.

What’s more, one of the areas where I think large LLMs are VERY useful is in the successful trawling of such public taxonomies that can, in turn, be referenced from within an LLM itself. For instance, using Claude, I can get a good sampling of IRIs for Barack Obama:

General Knowledge Bases

Wikidata
- IRI: http://www.wikidata.org/entity/Q76
- The canonical linked data hub; links to most other identifiers
DBpedia
- IRI: http://dbpedia.org/resource/Barack_Obama
- Wikipedia-derived; English DBpedia
YAGO
- IRI: http://yago-knowledge.org/resource/Barack_Obama

Library & Authority Control

VIAF (Virtual International Authority File)
- IRI: http://viaf.org/viaf/52010985
- Clusters 50+ national library records
Library of Congress Name Authority
- IRI: http://id.loc.gov/authorities/names/n2008054754
GND (German National Library)
- IRI: https://d-nb.info/gnd/132522136
ISNI
- IRI: https://isni.org/isni/0000000138447099
BnF (Bibliothèque nationale de France)
- IRI: https://data.bnf.fr/ark:/12148/cb15505097z
SUDOC (French academic libraries)
- IRI: https://www.idref.fr/119585049

Academic & Research

ORCID
- Not applicable (ORCID is for researchers claiming their own works)
Freebase (archived, but still referenced)
- ID: /m/02mjmr
- Google Knowledge Graph derived from this

Media & Entertainment

IMDb
- ID: nm1682433
- No official RDF; access via Wikidata property P345
MusicBrainz
- IRI: https://musicbrainz.org/artist/85752a4a-c3d3-4c5c-9a5a-6a6c82d4b921
- Listed as artist (Grammy-winning audiobook narrations)

Geographic (birthplace reference)

GeoNames (Honolulu)
- IRI: https://sws.geonames.org/5856195/
- His birthplace; linked from Wikidata P19

Social & Web

Twitter/X
- Username: BarackObama
- Wikidata property P2002
Facebook
- ID: barackobama
- Wikidata property P2013

Similarly, if I ask:

Give me a description in Turtle RDF, using the schema.org namespace, for the entity with IRI: http://www.wikidata.org/entity/Q76

I get the response:

@prefix schema: <http://schema.org/> .
@prefix wd: <http://www.wikidata.org/entity/> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .

wd:Q76 a schema:Person ;
    schema:name "Barack Obama"@en ;
    schema:givenName "Barack" ;
    schema:familyName "Obama" ;
    schema:additionalName "Barack Hussein Obama II" ;
    schema:description "44th President of the United States (2009–2017)"@en ;
    schema:birthDate "1961-08-04"^^xsd:date ;
    schema:birthPlace wd:Q18094 ;  # Honolulu
    schema:gender schema:Male ;
    schema:nationality wd:Q30 ;  # United States
    schema:spouse wd:Q13133 ;  # Michelle Obama
    schema:children wd:Q15070044 , wd:Q15070048 ;  # Malia, Sasha
    schema:alumniOf wd:Q49088 ,   # Columbia University
                    wd:Q13371 ;   # Harvard Law School
    schema:hasOccupation [
        a schema:Occupation ;
        schema:name "politician"
    ] , [
        a schema:Occupation ;
        schema:name "lawyer"
    ] , [
        a schema:Occupation ;
        schema:name "author"
    ] ;
    schema:jobTitle "44th President of the United States" ;
    schema:worksFor wd:Q35535 ;  # United Nations (post-presidency affiliations)
    schema:memberOf wd:Q29552 ;  # Democratic Party
    schema:award wd:Q37922 ,     # Nobel Peace Prize
                 wd:Q41254 ;     # Grammy Award
    schema:sameAs <http://viaf.org/viaf/52010985> ,
                  <http://dbpedia.org/resource/Barack_Obama> ,
                  <http://id.loc.gov/authorities/names/n2008054754> ,
                  <https://d-nb.info/gnd/132522136> ,
                  <https://isni.org/isni/0000000138447099> ;
    schema:image <https://commons.wikimedia.org/wiki/Special:FilePath/President_Barack_Obama.jpg> ;
    schema:url <https://barackobama.com/> .

# Birthplace expansion
wd:Q18094 a schema:Place ;
    schema:name "Honolulu"@en ;
    schema:containedInPlace wd:Q782 .  # Hawaii

# Spouse expansion  
wd:Q13133 a schema:Person ;
    schema:name "Michelle Obama"@en .

# Educational institutions
wd:Q49088 a schema:CollegeOrUniversity ;
    schema:name "Columbia University"@en .

wd:Q13371 a schema:CollegeOrUniversity ;
    schema:name "Harvard Law School"@en .

# Political party
wd:Q29552 a schema:Organization ;
    schema:name "Democratic Party"@en .

# Awards
wd:Q37922 a schema:Thing ;
    schema:name "Nobel Peace Prize"@en .

wd:Q41254 a schema:Thing ;
    schema:name "Grammy Award"@en .

Note here that I’m using the LLM for two things - resolving publicly addressable IRIs (here, Barack Obama’s Wikidata IRI http://www.wikidata.org/entity/Q76), then converting the result of this into a known ontology (schema.org). Note that this doesn’t guarantee consistency - you may get different data (and the possibility of hallucinations) - but it is far superior to what has gone before.

There’s a key point from the Claude conversation that I also want to pass on:

This uses Wikidata IRIs as the subject and for related entities, but expresses the relationships using schema.org vocabulary. The schema:sameAs property links out to the equivalent IRIs in other authority systems, enabling traversal across the linked data web.

In other words, here is an example in which one IRI can traverse similar entities across different knowledge bases, which was the fundamental premise of Linked Data in the first place.

Note that this approach can be broken down as follows: identify one or more reference standards you want to use to retrieve relevant information (e.g., the Wikidata identifier for President Barack Obama), and determine whether that entity is defined in that system. If this entity cannot be resolved, return something like the following so that you have a clear indication that an item doesn’t exist in that schema along with potential resolvers:


[] rdfs:label “Barack Biden” ; 
   ex:resolvable false ; 
   schema:iri <http://dbpedia.org/> ;
   # IRIs for Barack Obama and Joe Biden .
   schema:alternativeIRI <http://dbpedia.org/entity/Q76>, 
                      <http://www.wikidata.org/entity/Q6279> ;
   .

Note that this approach won’t necessarily work well for private taxonomies such as a company’s stores, employees, or parts lists, and there is no guarantee that very recent entries will be in these systems. However, if the choice is between standing up a knowledge graph to hold common reference data vs. using a publicly derived, already incorporated taxonomy from an LLM (say for place names, which benefit immensely from this technique), use the LLM.

Also note, that this moves well beyond people. Consider Benzyl Propionate, a chemical additive. It has identifiers in several systems (per prompt “Provide the properties for Benzyl Propionate as Turtle” as per Claude Opus 4.5):

@prefix schema: <http://schema.org/> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix skos: <http://www.w3.org/2004/02/skos/core#> .
@prefix obo: <http://purl.obolibrary.org/obo/> .
@prefix cheminf: <http://semanticscience.org/resource/> .
@prefix pubchem: <http://rdf.ncbi.nlm.nih.gov/pubchem/compound/> .
@prefix chebi: <http://purl.obolibrary.org/obo/CHEBI_> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
@prefix owl: <http://www.w3.org/2002/07/owl#> .

pubchem:CID31219 a schema:MolecularEntity , obo:CHEBI_23367 ;
    rdfs:label "Benzyl propionate"@en ;
    skos:prefLabel "Benzyl propionate"@en ;
    skos:altLabel "Benzyl propanoate"@en ,
                  "Phenylmethyl propanoate"@en ,
                  "Propanoic acid, phenylmethyl ester"@en ,
                  "Propionic acid benzyl ester"@en ;
    schema:description "A carboxylic ester used as a flavoring agent with a sweet, fruity odor"@en ;
    
    # Chemical identifiers
    schema:identifier [
        a schema:PropertyValue ;
        schema:propertyID "CAS" ;
        schema:value "122-63-4"
    ] ;
    cheminf:CHEMINF_000399 "122-63-4" ;  # CAS Registry Number
    
    # Structural properties
    schema:molecularFormula "C10H12O2" ;
    cheminf:CHEMINF_000334 "C10H12O2" ;  # Molecular Formula
    
    schema:molecularWeight "164.20"^^xsd:decimal ;
    cheminf:CHEMINF_000338 "164.201"^^xsd:decimal ;  # Molecular Weight
    
    schema:monoisotopicMolecularWeight "164.083729624"^^xsd:decimal ;
    
    # InChI and InChIKey
    schema:inChI "InChI=1S/C10H12O2/c1-2-10(11)12-8-9-6-4-3-5-7-9/h3-7H,2,8H2,1H3" ;
    schema:inChIKey "VHOMAPWVLKRQAZ-UHFFFAOYSA-N" ;
    
    # SMILES representations
    cheminf:CHEMINF_000018 "CCC(=O)OCc1ccccc1" ;  # Canonical SMILES
    cheminf:CHEMINF_000379 "CCC(=O)OCC1=CC=CC=C1" ;  # Isomeric SMILES
    
    # Physical properties
    schema:meltingPoint [
        a schema:PropertyValue ;
        schema:value "-30"^^xsd:decimal ;
        schema:unitCode "CEL"
    ] ;
    schema:boilingPoint [
        a schema:PropertyValue ;
        schema:value "222"^^xsd:decimal ;
        schema:unitCode "CEL"
    ] ;
    cheminf:CHEMINF_000454 "1.028"^^xsd:decimal ;  # Density (g/cm³)
    
    # Solubility
    schema:solubility "Soluble in alcohol and most fixed oils; slightly soluble in propylene glycol; insoluble in water and glycerin"@en ;
    
    # Classification
    obo:IAO_0000115 "A carboxylic ester that is the propanoate ester of benzyl alcohol"@en ;  # textual definition
    rdfs:subClassOf obo:CHEBI_33308 ;  # carboxylic ester
    obo:RO_0000087 obo:CHEBI_48318 ;  # has role: flavouring agent
    
    # Functional use
    schema:category "Flavoring agent"@en , "Fragrance ingredient"@en ;
    
    # Organoleptic properties
    schema:additionalProperty [
        a schema:PropertyValue ;
        schema:propertyID "odor" ;
        schema:value "Sweet, fruity, floral"
    ] , [
        a schema:PropertyValue ;
        schema:propertyID "taste" ;
        schema:value "Fruity, sweet, green, powdery with ripe berry nuance (at 15 ppm)"
    ] ;
    
    # Safety classification
    schema:additionalProperty [
        a schema:PropertyValue ;
        schema:propertyID "GHS_classification" ;
        schema:value "Not classified as hazardous"
    ] ;
    
    # Regulatory status
    schema:additionalProperty [
        a schema:PropertyValue ;
        schema:propertyID "FEMA_GRAS" ;
        schema:value "2150"
    ] , [
        a schema:PropertyValue ;
        schema:propertyID "FDA_status" ;
        schema:value "Generally Recognized as Safe (GRAS)"
    ] ;
    
    # Linked identifiers (sameAs)
    owl:sameAs chebi:41492 ,
               <http://www.chemspider.com/Chemical-Structure.28960.html> ,
               <http://www.wikidata.org/entity/Q27251695> ;
    
    # Provenance
    schema:isPartOf <http://rdf.ncbi.nlm.nih.gov/pubchem/> ;
    rdfs:seeAlso <https://pubchem.ncbi.nlm.nih.gov/compound/31219> ,
                 <https://www.chemspider.com/Chemical-Structure.28960.html> .

# Related substance: Benzyl alcohol (precursor)
<http://rdf.ncbi.nlm.nih.gov/pubchem/compound/CID244> a schema:MolecularEntity ;
    rdfs:label "Benzyl alcohol"@en ;
    schema:molecularFormula "C7H8O" .

# Related substance: Propionic acid (precursor)
<http://rdf.ncbi.nlm.nih.gov/pubchem/compound/CID1032> a schema:MolecularEntity ;
    rdfs:label "Propionic acid"@en ;
    schema:molecularFormula "C3H6O2" .

# Synthesis relationship
pubchem:CID31219 schema:isBasedOn <http://rdf.ncbi.nlm.nih.gov/pubchem/compound/CID244> ,
                                  <http://rdf.ncbi.nlm.nih.gov/pubchem/compound/CID1032> .

Note that this goes the other way as well. For instance, if you have content that you frequently utilize (such as states in the US or provinces in Canada), you can retrieve not only the data but also relevant public IRIs via a single LLM query that can in turn be stored in your knowledge graph.
The query:

Using the Geonames taxonomy, generate a Turtle RDF showing the 50 states of the United States, including relevant IRIs.

The output:

# Turtle RDF showing (some of) the 50 states of the USA.
@prefix gn: <http://www.geonames.org/ontology#> .
@prefix geo: <http://www.w3.org/2003/01/geo/wgs84_pos#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix skos: <http://www.w3.org/2004/02/skos/core#> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
@prefix owl: <http://www.w3.org/2002/07/owl#> .
@prefix dcterms: <http://purl.org/dc/terms/> .
@prefix wd: <http://www.wikidata.org/entity/> .

# United States of America (parent country)
<https://sws.geonames.org/6252001/> a gn:Feature ; # Geonames link
    rdfs:label "United States"@en ;
    gn:name "United States" ;
    gn:featureClass gn:A ;
    gn:featureCode gn:A.PCLI ;
    gn:countryCode "US" ;
    owl:sameAs wd:Q30 . # Wikidata link. 

# ============================================================================
# The 50 States of the United States
# GeoNames Feature Code: A.ADM1 (first-order administrative division)
# ============================================================================

# Alabama
<https://sws.geonames.org/4829764/> a gn:Feature ; # Geonames link
    rdfs:label "Alabama"@en ;
    gn:name "Alabama" ;
    gn:alternateName "AL" ;
    gn:featureClass gn:A ;
    gn:featureCode gn:A.ADM1 ;
    gn:countryCode "US" ;
    gn:adminCode1 "AL" ;
    gn:parentCountry <https://sws.geonames.org/6252001/> ;
    geo:lat "32.75041"^^xsd:decimal ;
    geo:long "-86.75026"^^xsd:decimal ;
    owl:sameAs wd:Q173 . # Wikidata link. 

# Alaska
<https://sws.geonames.org/5879092/> a gn:Feature ; # Geonames link
    rdfs:label "Alaska"@en ;
    gn:name "Alaska" ;
    gn:alternateName "AK" ;
    gn:featureClass gn:A ;
    gn:featureCode gn:A.ADM1 ;
    gn:countryCode "US" ;
    gn:adminCode1 "AK" ;
    gn:parentCountry <https://sws.geonames.org/6252001/> ;
    geo:lat "64.00028"^^xsd:decimal ;
    geo:long "-150.00028"^^xsd:decimal ;
    owl:sameAs wd:Q797 . # Wikidata link.

The output can be readily reconfigured for canonical internal ontologies, but the idea of storing IRIs as references in knowledge graphs to known public entities has significant merit.

Public Taxonomies

Here’s a basic guide (as per Claude Opus 4.5) for standard taxonomies that I keep handy.

General-Purpose Knowledge Bases

Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Most comprehensive coverage; Q-identifiers; SPARQL endpoint; excellent for LLMs due to multilingual labels and descriptions
DBpedia
- Base IRI: http://dbpedia.org/resource/
- Wikipedia-derived; stable URIs; multiple language chapters
YAGO
- Base IRI: http://yago-knowledge.org/resource/
- Wikipedia + WordNet derived; strong typing

People

VIAF (Virtual International Authority File)
- Base IRI: http://viaf.org/viaf/
- Coverage: Authors, historical figures; clusters national authority files
ISNI (International Standard Name Identifier)
- Base IRI: https://isni.org/isni/
- Coverage: Creators and contributors broadly
ORCID
- Base IRI: https://orcid.org/
- Coverage: Living researchers and academics
LC Name Authority (Library of Congress)
- Base IRI: http://id.loc.gov/authorities/names/
- Coverage: Comprehensive; US bias but international coverage
GND (German National Library)
- Base IRI: https://d-nb.info/gnd/
- Coverage: Strong European coverage
Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Coverage: Contemporary, historical, and fictional persons via P31 = Q5 or subclasses

For fictional people, Wikidata is the most systematic source (P31 = Q95074), with good coverage of literary, film, and game characters.

Places

GeoNames
- Base IRI: https://sws.geonames.org/
- ~12M features; CC-BY license; stable numeric IDs
Getty TGN (Thesaurus of Geographic Names)
- Base IRI: http://vocab.getty.edu/tgn/
- Historical and current places; art-world focus
Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Linked to GeoNames, TGN, ISO codes
ISO 3166 (via Library of Congress)
- Base IRI: http://id.loc.gov/vocabulary/countries/
- Countries and subdivisions
LinkedGeoData
- Base IRI: http://linkedgeodata.org/triplify/
- OpenStreetMap as RDF

For fictional places, Wikidata provides the most systematic coverage with appropriate typing.

Organizations

LEI (Legal Entity Identifier via GLEIF)
- Base IRI: https://rdf.gleif.org/L1/
- Legal entities in finance; ISO 17442 standard
ROR (Research Organization Registry)
- Base IRI: https://ror.org/
- Research organizations; ~100K entities
Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Broad organizational coverage
ISNI
- Base IRI: https://isni.org/isni/
- Also covers organizations, not just individuals
Grid.ac
- Base IRI: https://www.grid.ac/institutes/
- Research institutions (now feeding into ROR)

Books and Media Works

Open Library
- Base IRI: https://openlibrary.org/works/
- Books; linked to ISBN
OCLC/WorldCat
- Base IRI: http://www.worldcat.org/oclc/
- Library holdings worldwide
Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Works and editions with VIAF/ISBN links
MusicBrainz
- Base IRI: https://musicbrainz.org/
- Music works, recordings, artists
EIDR (Entertainment Identifier Registry)
- Base IRI: https://ui.eidr.org/view/content?id=
- Entertainment works; ISO standard
IMDB
- No open IRIs; use Wikidata with property P345 for cross-reference

Chemicals, Elements, Substances

PubChem
- Base IRI: http://rdf.ncbi.nlm.nih.gov/pubchem/compound/
- ~100M compounds; RDF available
ChEBI (Chemical Entities of Biological Interest)
- Base IRI: http://purl.obolibrary.org/obo/CHEBI_
- Biological interest chemicals; OBO ontology
ChEMBL
- Base IRI: https://www.ebi.ac.uk/chembl/compound_report_card/
- Bioactive molecules
Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Elements and compounds with InChI, CAS links
IUPAC Gold Book
- Base IRI: https://goldbook.iupac.org/terms/view/
- Chemical terminology definitions

For elements, the periodic table is well-covered in Wikidata with links to PubChem and ChEBI.

Biology: Animals and Plants

NCBI Taxonomy
- Base IRI: http://purl.obolibrary.org/obo/NCBITaxon_
- Comprehensive; OBO format
GBIF (Global Biodiversity Information Facility)
- Base IRI: https://www.gbif.org/species/
- Global biodiversity; occurrence data
ITIS (Integrated Taxonomic Information System)
- Base IRI: https://www.itis.gov/servlet/SingleRpt/SingleRpt?search_topic=TSN&search_value=
- North American focus
UniProt Taxonomy
- Base IRI: https://purl.uniprot.org/taxonomy/
- Aligned with NCBI
EOL (Encyclopedia of Life)
- Base IRI: https://eol.org/pages/
- General species information
IPNI (International Plant Names Index)
- Base IRI: https://www.ipni.org/n/
- Plant nomenclature
WFO (World Flora Online)
- Base IRI: http://www.worldfloraonline.org/taxon/wfo-
- Authoritative plant taxonomy
POWO (Plants of the World Online)
- Base IRI: https://powo.science.kew.org/taxon/urn:lsid:ipni.org:names:
- Kew Gardens reference

Legal Terminology

EuroVoc
- Base IRI: http://eurovoc.europa.eu/
- EU multilingual thesaurus
FIBO (Financial Industry Business Ontology)
- Base IRI: https://spec.edmcouncil.org/fibo/ontology/
- Financial and legal ontology
LC Subject Headings
- Base IRI: http://id.loc.gov/authorities/subjects/
- Includes legal topics
Akoma Ntoso
- Base IRI: http://docs.oasis-open.org/legaldocml/ns/akn/3.0/
- Legal document markup standard
Wikidata
- Use property P31 = Q7748 for legal concepts

Physical Constants and Units

QUDT (Quantities, Units, Dimensions, Types)
- Units: http://qudt.org/vocab/unit/
- Constants: http://qudt.org/vocab/constant/
- Excellent RDF coverage
OM (Ontology of Units of Measure)
- Base IRI: http://www.ontology-of-units-of-measure.org/resource/om-2/
- Alternative units ontology
CODATA
- Available via Wikidata properties; linked from QUDT
UCUM (Unified Code for Units of Measure)
- Base IRI: http://unitsofmeasure.org/
- Healthcare and scientific focus

Mathematical Concepts

Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Theorems, algorithms, mathematicians
DBpedia
- Base IRI: http://dbpedia.org/resource/
- Good coverage of named theorems
OEIS (Online Encyclopedia of Integer Sequences)
- Base IRI: https://oeis.org/
- Integer sequences with identifiers
nLab
- Base IRI: https://ncatlab.org/nlab/show/
- Category theory focus
MathWorld
- No stable IRIs; reference only

Historical Events

Wikidata
- Base IRI: http://www.wikidata.org/entity/
- Best coverage; use P31 = Q1190554 (occurrence)
DBpedia
- Base IRI: http://dbpedia.org/resource/
- Events, battles, treaties
PeriodO
- Base IRI: http://n2t.net/ark:/99152/
- Historical periods and cultural eras
Getty AAT (Art & Architecture Thesaurus)
- Base IRI: http://vocab.getty.edu/aat/
- Cultural/art events and concepts
CIDOC-CRM
- Base IRI: http://www.cidoc-crm.org/cidoc-crm/
- Event modeling ontology for cultural heritage

Additional Taxonomic Categories

Occupations — ESCO
- Base IRI: http://data.europa.eu/esco/occupation/
Diseases — SNOMED CT
- Base IRI: http://purl.bioontology.org/ontology/SNOMEDCT/
Drugs — DrugBank
- Base IRI: http://wifo5-03.informatik.uni-mannheim.de/drugbank/resource/drugs/
Foods — FoodOn
- Base IRI: http://purl.obolibrary.org/obo/FOODON_
Languages — Glottolog
- Base IRI: https://glottolog.org/resource/languoid/id/
Currencies — ISO 4217 via Wikidata
- Base IRI: http://www.wikidata.org/entity/
Art/Cultural Objects — Getty AAT
- Base IRI: http://vocab.getty.edu/aat/
Gene Ontology — GO
- Base IRI: http://purl.obolibrary.org/obo/GO_
Anatomy — Uberon
- Base IRI: http://purl.obolibrary.org/obo/UBERON_

Practical Notes for LLM Integration

Best overall sources for LLM grounding:

Wikidata — broadest coverage, multilingual labels/descriptions, active SPARQL endpoint, consistent Q-identifier pattern
Library of Congress — authoritative for names, subjects, languages; stable id.loc.gov URIs
OBO Foundry ontologies — biological/chemical domains; consistent purl.obolibrary.org pattern
QUDT — physical sciences; well-documented

Key considerations:

Prefer sources with rdfs:label and rdfs:comment or skos:definition — essential for LLM context
Wikidata’s schema:description property is particularly useful for disambiguation
Many OBO ontologies include obo:IAO_0000115 (textual definition) which works well for prompts
For reconciliation, OpenRefine reconciliation endpoints exist for Wikidata, VIAF, GeoNames, and others

Summary

As an ontologist, taxonomist, or information architect, keep an eye out for such public repositories. Whenever you are ingesting data into your knowledge graph, part of your ingestion pipeline should include querying against an LLM to determine whether the concept that you are putting into the knowledge graph doesn’t already have one or more public IRIs. If it does, add them in (schema:isBasedOn or owl:sameAs are both useful here). Such IRIs work exceptionally well when integrating public taxonomies via LLMs and can significantly reduce the complexity of master data management, especially for frequently referenced concepts.

Also, if publicly available industry taxonomies exist, LLMs are a good place to query about them. No one can be a subject-matter expert in everything. Moreover, even when there isn’t a comprehensive taxonomy (e.g., Health Insurance is a mess), such a search can identify potential areas for curation to help create or enrich taxonomies that expose public IRIs.

Disclosure: portions of this article were prepared with Claude Opus 4.5.

In media res,

Kurt Cagle
The Ontologist

Check out my LinkedIn newsletter, The Cagle Report.

I am also currently seeking new projects or work opportunities. If anyone is looking for a CTO or Director-level AI/Ontologist, please get in touch with me through my Calendly:

If you want to shoot the breeze or have a cup of virtual coffee, I have a Calendly account at https://calendly.com/theCagleReport. I am available for consulting and full-time work as an ontologist, AI/Knowledge Graph guru, and coffee maker. Also, for those of you whom I have promised follow-up material, it’s coming; I’ve been dealing with health issues of late.

I’ve created a Ko-fi account for voluntary contributions, either one-time or ongoing, or you can subscribe directly to The Ontologist. If you value my articles, technical pieces, or general reflections on work in the 21st century, please consider contributing to support my work and allow me to continue writing.

The Ontologist

Discussion about this post

Ready for more?