Figure 6

Normalization of a composite entity. Starting with the text “differences in coumadin requirements”, NLP tools generate the raw entity “coumadin requirements differences” on which we can apply the normalization algorithm (described in table 1) using the PHARE ontology. The first step ensures that the preferred name warfarin is used instead of coumadin. The second step maps “requirements” to the entity type DrugDose, and the final step maps “differences” to the entity type Variation. The axiom noted with a * is added to the ontology during the normalization as a result of the inference that a variation in drug dose was found.