Grouping Axioms for More Coherent Ontology Descriptions

Ontologies and datasets for the Semantic Web are encoded in OWL formalisms that are not easily comprehended by people. To make ontologies accessible to human domain experts, several research groups have developed ontology verbalisers using Natural Language Generation. In practice ontologies are usually composed of simple axioms, so that realising them separately is relatively easy; there remains however the problem of producing texts that are coherent and efficient. We describe in this paper some methods for producing sentences that aggregate over sets of axioms that share the same logical structure. Because these methods are based on logical structure rather than domain-specific concepts or language-specific syntax, they are generic both as regards domain and language.