Terminology standards at work: exploring several scenarios to implement ISO 16642

Colloque sur invitation. internationale.International audienceThere is always some distance between the technical constraints that are expressed in an international standard and the actual usage an implementer can make out of them in the practical settings of its company or institution. This is all the more true for ISO 16642 (Computer applications in Terminology Terminology Markup Framework), which does not provide one single computer format for representing computerized terminologies, but rather focuses on the underlying model that any such format should be conformant with. As a consequence, it is essential to make further studies on how specific usage scenarios may lead to even more constraints on the possible data models, so that general trends can be identified and, possibly, made explicit as recommendations to the final-end users. One possible format for such recommendations will of course be the identification of data category selections (DCS) that may characterize families of interoperable data structures on the basis of the set of data categories made available by ISO standard 12620 (Computer applications in Terminology Data Categories). In this context, our paper will explore three particular situations that will be qualified according to the condition of production or consumption of terminological data, and accordingly for which the core of a possible standard DCS will be described. The three situations are the following ones: - the unification of several thesauri and term banks within a nationwide scientific documentation center; - the initial deployment of a terminology component within a large industrial group; - the definition of a pivot representation for terminology extraction components that can transparently articulate the output of such software with higher level knowledge management components. As we shall see, the flexibility offered by ISO 16642 as well as its sound principles allow us to carry such specification tasks without precluding any further evolution in those various applicative contexts. We will also see how the methodology outlined in this paper allows us to provide wider information architectures combining two or more of the scenarios that we have identified