How to evaluate necessary cooperative systems of terminology building?
Proceedings of the Third International Conference on Language Resources and Evaluation (LREC 2002)
Abstract
Terminology building cannot be considered as a full automated process but rather as a cooperative task between terminological tools and terminologists. Identifying terms in a technical domain is a matter of word usage and expert agreement. We point out the problem of the evaluation of such tools: their quality and their contribution to the terminology building is difficult to estimate and cannot be fully evaluated with usual precision and recall measures. We aim at evaluating more globally their technical aspects and their usability. We give a non-exhaustive list of the features of such evaluation. Then, we apply them on four terminological systems.