OAEI Bio-ML 2026

Subtrack 1 — Submission Format

One OAEI Alignment RDF file per pair, with full OWL IRIs — the exact cell shape, a worked example, and a copy-paste template.

Subtrack 1 of Track 1 is a whole-ontology equivalence alignment over the three Bio-ML 2026 pairs:

  • NCIT-DOID
  • SNOMED-FMA
  • SNOMED-NCIT

For each pair you submit one complete alignment in the OAEI Alignment RDF format, using full OWL IRIs. Submissions are scored organiser-side against the hidden test reference — repaired, coherence-aware P/R/F1 (headline) plus standard P/R/F1 and Global Coherence (see the task description).

File format

Root is rdf:RDF. The default namespace (xmlns=) is the OAEI Alignment namespace without a trailing #: http://knowledgeweb.semanticweb.org/heterogeneity/alignment. Exactly one <Alignment> with a header (<xml>yes</xml>, <level>0</level>, <type>??</type>, and <onto1> / <onto2> URIs). Each correspondence is a <map> wrapping one <Cell>:

  • <entity1 rdf:resource="IRI"/> — the source ontology entity’s full OWL IRI,
  • <entity2 rdf:resource="IRI"/> — the target ontology entity’s full OWL IRI,
  • <relation>=</relation> — equivalence (Subtrack 1 evaluates equivalence),
  • <measure rdf:datatype="http://www.w3.org/2001/XMLSchema#float">1.0</measure> — confidence in [0, 1].

Submit one file per pair. Entity IRIs MUST be absolute — this track uses full OWL IRIs, not the CURIEs used by Track 2. Orientation matters: <entity1> must come from onto1 (source) and <entity2> from onto2 (target); a reversed cell can never match the reference. The order of <map> elements is irrelevant. Submit a full alignment over the two ontologies; the organisers score the portion that lands on the hidden test entities.

Worked example (illustrative)

The cell below is illustrative — it shows the exact shape, using full OWL IRIs of the kind found in NCIT-DOID:

<?xml version="1.0" encoding="utf-8"?>
<rdf:RDF xmlns="http://knowledgeweb.semanticweb.org/heterogeneity/alignment"
         xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
         xmlns:xsd="http://www.w3.org/2001/XMLSchema#">
  <Alignment>
    <xml>yes</xml>
    <level>0</level>
    <type>??</type>
    <onto1>http://ncicb.nci.nih.gov/xml/owl/EVS/Thesaurus.owl</onto1>
    <onto2>http://purl.obolibrary.org/obo/doid.owl</onto2>
    <map>
      <Cell>
        <entity1 rdf:resource="http://ncicb.nci.nih.gov/xml/owl/EVS/Thesaurus.owl#C3262"/>
        <entity2 rdf:resource="http://purl.obolibrary.org/obo/DOID_162"/>
        <relation>=</relation>
        <measure rdf:datatype="http://www.w3.org/2001/XMLSchema#float">1.0</measure>
      </Cell>
    </map>
  </Alignment>
</rdf:RDF>

Minimal copy-paste template

<?xml version="1.0" encoding="utf-8"?>
<rdf:RDF xmlns="http://knowledgeweb.semanticweb.org/heterogeneity/alignment"
         xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
         xmlns:xsd="http://www.w3.org/2001/XMLSchema#">
  <Alignment>
    <xml>yes</xml>
    <level>0</level>
    <type>??</type>
    <onto1>SOURCE_ONTOLOGY_URI</onto1>
    <onto2>TARGET_ONTOLOGY_URI</onto2>
    <map>
      <Cell>
        <entity1 rdf:resource="SOURCE_IRI"/>
        <entity2 rdf:resource="TARGET_IRI"/>
        <relation>=</relation>
        <measure rdf:datatype="http://www.w3.org/2001/XMLSchema#float">1.0</measure>
      </Cell>
    </map>
  </Alignment>
</rdf:RDF>

Validate — scoring is organiser-side

Subtrack 1 is semi-supervised: the test reference is hidden, so you cannot self-score against it. Validate the structure of each alignment file with the kit’s validate_global.py (Python 3.12+, standard library; it also checks against the bundled RelaxNG schema alignment.rng), then submit:

# structural + schema validation of one alignment file
python scoring_kit/validate_global.py  my-ncit-doid.rdf

To sanity-check your matcher before submitting, compare its output against the public refs_equiv/train.tsv correspondences (a proxy — there is no public global scorer, and the hidden test slice is scored organiser-side). The organisers then compute the headline repaired, coherence-aware P/R/F1, the standard P/R/F1, and Global Coherence, and publish them to the Track 1 — Global Alignment leaderboard.