Enriching the Human Phenotype Ontology with Inferred Axioms from Textual Descriptions.

The Human Phenotype Ontology (HP) is a reference vocabulary of human phenotypic abnormalities. HP, apart from the textual information (general definitions, descriptions, synonyms, etc.) of each ontology concept, also provides computer-readable logical definitions (axioms) of terms that will allow human phenotypic abnormalities to be related to entities from anatomy, pathology, biochemistry and other areas. In this paper we present a prototype to generate new axiomatic knowledge from the textual descriptions of each HP term. The prototype (i) detects terms in the textual descriptions and not found in the given logical expressions, (ii) generates pair combinations of those terms, (iii) builds triples after detecting the most probable relation between the pair of terms using a statistical model and, finally, (iv) suggests the most probable triples to the user so she can decide which ones can be added to the original axioms.