V. Claveau > Publications > TKE'05 Abstract OLST                                    Vincent Claveau and Marie-Claude L'Homme Structuring Terminology by Analogy-Based Machine Learning, 7th International Conference on Terminology and Knowledge Engineering, (TKE'05), Copenhagen, Denmark, August 2005, Document (pdf) Abstract This paper explores the way a simple machine learning technique together with a terminological extraction system can be used to find whether a term is related to another. Our work bears a number of similarities with that developed by Zweigenbaum & Grabar (2003), but it also aims at precisely predicting the semantic link between the two terms. This work relies on two main hypotheses:    1. specialized corpora contain regular morphological relationships coupled with a regular semantic relation;    2. such morphological links may be “exclusive” to the studied domain. The machine learning technique based on analogy we propose allows us to take into account the particularities of this classification task and does not need any external morphology knowledge in order to comply with our second hypothesis. The whole technique is evaluated in the domain of computer science and applied on a French corpus.