Ontology Management System

For linguistics we can envisage two kinds of interoperability. The first kind facilitates data sharing. The second, more ambitious, kind provides the basis for comparison and checking of theoretical claims about languages. Unfortunately, interoperability of the first kind does not guarantee interoperability of the second kind, because the intuitions of linguists, even when they have used an interoperable format to describe a language, may not be recoverable. If one wishes to take the more ambitious route and develop interoperability of the second kind, a major problem faced is that linguists often do not agree about the category or type to which a particular phenomenon belongs. The reasons for this are numerous, but often it has to do with choosing a particular characteristic or property as fundamental or defining. To illustrate with a concrete example from Russian, emphasis can be placed on a number of different properties when defining the notion ‘agreement’, including that the controller and target of agreement mark the same feature value. But this might lead one to exclude examples such as the following, when similar constructions in the language would fit: