Sometimes in AI we encounter overlapping usage of the terms ontology and Knowledge Base (KB): an ontology includes a taxonomic hierarchy, which determines the inheritance, plus other facts about objects, relations, events and individuals. So in some approaches the ontology can be called a KB (and it is called in this way, at least by some researchers). In this paper we will distinguish the two notions: ontology is "a particular theory of the nature of being or existence. It determines what kinds of things exist, but does not determine their specific properties and interrelationships" [Russell & Norvig 95]. An ontology represents knowledge that does not change with inference, it is supposed to be stable (but open to monotonic enhancement) and true in all possible worlds which are relevant for a certain community of users [Guarino & Giaretta 95]. On the other hand, a KB is a set of facts about the world and includes an ontology; the KB contains facts about individuals which are true under certain constraints or in some contexts as well as axioms providing the possible interpretations. Usually the notions of inference and reasoning are related to a KB.
Regardless of the adopted terminology, most knowledge-based NLP systems rely on knowledge resources ranging between the above definitions of ontology and KBs (there is now a consensus that the taxonomy itself is rather insufficient for NLP). These resources contain reusable ontological content which is often influenced by ( i) the specific task; ( ii) natural languages used for knowledge acquisition; ( iii) particularities of the representation formalism; and ( iv) specific inference procedures performed by the system.