The two evaluation tracks of the benchmark. Each task takes a set of competency questions as input and produces a different artefact.
| Short Name | Title | Coverage | Identifier | Keywords |
|---|---|---|---|---|
| CQ2Term | Term-level generation of classes and properties from competency questions | 2 dimensions | https://w3id.org/cq4oe/task/CQ2Term | Term Generation, Task, Competency Questions |
| CQ2Onto | Full OWL ontology generation from competency questions | 6 dimensions | https://w3id.org/cq4oe/task/CQ2Onto | Ontology Generation, Task, Competency Questions |
Each dimension defines a scoped target against which the atomic metrics are computed.
| Short Name | Title | Task | Identifier | Keywords |
|---|---|---|---|---|
| ClassProperty | Class and property generation | CQ2Term, CQ2Onto | https://w3id.org/cq4oe/dimension/ClassProperty | Term Generation, Ontology Generation, Dimension, Classes, Properties |
| PropertyCharacteristics | Property characteristics modelling | CQ2Onto | https://w3id.org/cq4oe/dimension/PropertyCharacteristics | Property Modelling, Dimension |
| Triple | Domain and range triple evaluation | CQ2Onto | https://w3id.org/cq4oe/dimension/Triple | Domain/Range, Dimension, Triples |
| Axiom | TBox axiom-level structural evaluation | CQ2Onto | https://w3id.org/cq4oe/dimension/Axiom | TBox Axioms, Dimension |
| HierarchyClosure | Hierarchy closure under reasoning | CQ2Onto | https://w3id.org/cq4oe/dimension/HierarchyClosure | Hierarchy Closure, Dimension |
| CQCoverage | CQ condition coverage under clouser level | CQ2Term, CQ2Onto | https://w3id.org/cq4oe/dimension/CQCoverage | Competency Questions, Coverage, Dimension |
The atomic measurements computed by the evaluation pipeline. Each metric is a single, reusable measurement that may be applied across one or more dimensions.
| Short Name | Title | Category | Dimension(s) | Identifier | Keywords | Status |
|---|---|---|---|---|---|---|
| HardMatch | Exact lexical string comparison | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/HardMatch | Ontology Concept Matching, Metric, Classes, Properties, Lexical | Active |
| SequenceMatch | Token-level string similarity comparison | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/SequenceMatch | Ontology Concept Matching, Metric, Classes, Properties, Lexical | Active |
| LevenshteinDistance | Character-level edit distance approximation | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/LevenshteinDistance | Ontology Concept Matching, Metric, Classes, Properties, Edit Distance | Active |
| JaroWinklerDistance | String similarity emphasizing common prefixes | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/JaroWinklerDistance | Ontology Concept Matching, Metric, Classes, Properties, Lexical | Active |
| SemanticCosineSimilarity | Semantic distance over dense vector representations | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/SemanticCosineSimilarity | Ontology Concept Matching, Metric, Classes, Properties, Encoder, Semantic | Active |
| AggregatedTop3 | Mean of the three highest non-hard similarity scores | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/AggregatedTop3 | Ontology Concept Matching, Metric, Score Aggregation | Active |
| Synonyms | Exact match against WordNet synonym relations | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/Synonyms | Ontology Concept Matching, Metric, Classes, Properties, WordNet | Inactive |
| InformationContent | Semantic-level comparison using WordNet synsets and IC scores | Ontology Concept Matching | ClassProperty | https://w3id.org/cq4oe/metric/InformationContent | Ontology Concept Matching, Metric, Classes, Properties, WordNet | Inactive |
| Precision | Fraction of generated elements that match the gold standard | Set-Based Evaluation | ClassProperty, PropertyCharacteristics, Triple, Axiom, HierarchyClosure | https://w3id.org/cq4oe/metric/Precision | Ontology Evaluation, Metric, Precision | Active |
| Recall | Fraction of gold elements matched by the generation | Set-Based Evaluation | ClassProperty, PropertyCharacteristics, Triple, Axiom, HierarchyClosure | https://w3id.org/cq4oe/metric/Recall | Ontology Evaluation, Metric, Recall | Active |
| F1 | Harmonic mean of precision and recall | Set-Based Evaluation | ClassProperty, PropertyCharacteristics, Triple, Axiom, HierarchyClosure | https://w3id.org/cq4oe/metric/F1 | Ontology Evaluation, Metric, F1 | Active |
| EmbeddingDiagnostic | Soft similarity over normalized textual serializations | Diagnostic Evaluation | Triple, Axiom | https://w3id.org/cq4oe/metric/EmbeddingDiagnostic | Ontology Evaluation, Diagnostic, Embedding, Triples, Axioms | Active |
| AtLeastOneCoverage | Fraction of CQs with at least one required element correctly generated | Competency Question Coverage | CQCoverage | https://w3id.org/cq4oe/metric/AtLeastOneCoverage | Ontology Engineering, Metric, Competency Questions, Coverage | Active |
| MeanCoverage | Average per-CQ coverage rate across all retained CQs | Competency Question Coverage | CQCoverage | https://w3id.org/cq4oe/metric/MeanCoverage | Ontology Engineering, Metric, Competency Questions, Coverage | Active |
| FullCoverage | Fraction of CQs whose required elements are fully generated | Competency Question Coverage | CQCoverage | https://w3id.org/cq4oe/metric/FullCoverage | Ontology Engineering, Metric, Competency Questions, Completeness, Coverage | Active |
| ClosureRescueCoverage | CQ coverage credited via reasoner-inferred subsumption closure | Competency Question Coverage | CQCoverage, HierarchyClosure | https://w3id.org/cq4oe/metric/ClosureRescueCoverage | Ontology Engineering, Metric, Competency Questions, Reasoning, HermiT, Closure | Active |