Applying Chaining Rules
SSSOM Mapping Chains
The goal of this document is to capture all obvious mapping chaining rules that could be applied to SSSOM,
and later delivered as part of sssom toolkit.
This is all structural, and should not be confused with proper reasoning or mapping reconciliation ala
boomer.
The idea is to provide the functionality to apply these chaining rules over a given mapping set, and record the appropriate metadata for that rule.
Rules:
Transitivity Rule
Transitivity of a relation R implies that if an entity A is R-related to an entity B which in turn is
R-related to an entity C, A is also R-related to C.
Predicates applicable in transitivity rules
We consider the following predicates transitive:
- skos:exactMatch
- skos:narrowMatch
- skos:broadMatch
- owl:equivalentClass / owl:equivalentProperty
- rdfs:subClassOf / rdfs:subPropertyOf
- owl:sameAs
Note that technically speaking skos:narrowMatch and skos:broadMatch are not considered transitive
(skos:broaderTransitive would be), but we are not defining a new semantics here,
just a reasonable default for a mapping tool, which will nearly always hold true.
Predicates we do not consider transitive include: skos:relatedMatch (for practical reasons), oboInOwl:hasDbXref,
skos:closeMatch, rdfs:seeAlso (weakest form of a mapping link), rdf:type.
Rules
- T1:
(:A)-[predicate_id]->(:B)-[predicate_id]->(:C)->(:A)-[predicate_id]->(:C)
Examples
- T1-EX:
(:A)-[skos:broadMatch]->(:B)-[skos:broadMatch]->(:C)->(:A)-[skos:broadMatch]->(:C)
Role chains over exact/equivalent matches
Role chains are rules that allow us to bridge across mappings across multiple different properties. Role chains over exact are simple to define, so we start with these
Predicates applicable in transitity rules
- skos:narrowMatch
- skos:broadMatch
- skos:closeMatch
- skos:relatedMatch
Rules for SKOS
- RCE1:
(:A)-[skos:exactMatch|owl:equivalentClass]->(:B)-[predicate_id]->(:C)->(:A)-[predicate_id]->(:C) - RCE2:
(:A)-[predicate_id]->(:B)-[skos:exactMatch]->(:C)->(:A)-[predicate_id]->(:C)
Rules that should probably not be inferred (OWL)
The following rules hold true, but will be left to a reasoner to be inferred:
- RCE-N1:
(:A)-[owl:equivalentClass]->(:B)-[rdfs:subClassOf]->(:C)->(:A)-[rdfs:subClassOf]->(:C) - RCE-N2:
(:A)-[rdfs:subClassOf]->(:B)-[owl:equivalentClass]->(:C)->(:A)-[rdfs:subClassOf]->(:C) - RCE-N3:
(:A)-[owl:equivalentProperty]->(:B)-[rdfs:subPropertyOf]->(:C)->(:A)-[rdfs:subPropertyOf]->(:C) - RCE-N4:
(:A)-[rdfs:subPropertyOf]->(:B)-[owl:equivalentProperty]->(:C)->(:A)-[rdfs:subPropertyOf]->(:C)
Inverse Rules
R inverse of S implies that if an entity A is R-related to an entity B then B is also S-related to A.
We like to call the output of an inverse rule a walk-back. A command that applies an inverse rule could be called flip.
Predicates applicable in inverse rules
This excludes the exact predicates for which inverse rules are redundant.
Rules for SKOS
- RI1:
(:A)-[skos:narrowMatch]->(:B)->(:B)-[skos:broadMatch]->(:A) - RI2:
(:A)-[skos:broadMatch]->(:B)->(:B)-[skos:narrowMatch]->(:A)
Rules for SEMAPV
- RI3:
(:A)-[semapv:crossSpeciesExactMatch]->(:B)->(:B)-[semapv:crossSpeciesExactMatch]->(:A) - RI4:
(:A)-[semapv:crossSpeciesNarrowMatch]->(:B)->(:B)-[semapv:crossSpeciesBroadMatch]->(:A) - RI5:
(:A)-[semapv:crossSpeciesBroadMatch]->(:B)->(:B)-[semapv:crossSpeciesNarrowMatch]->(:A)
Generalisation Rules
Generalisation rules are rules that can be applied to weaken a mapping deliberately. This is sometimes useful, for example when combining strong OWL-Semantics mappings with weaker SKOS-based ones.
Rules
- RG1:
(:A)-[owl:equivalentTo]->(:B)->(:A)-[skos:exactMatch]->(:B) - RG2:
(:A)-[owl:subClassOf]->(:B)->(:A)-[skos:broadMatch]->(:B)