Morpholog: Constrained and Supervised Learning of Morphology
Rémi Zajac · 2001
We present an automated method for building morphological analyzers using limited data elicited from a linguist. The system uses an induction procedure based on an extended edit distance algorithm to automate the discovery of morphological rules from examples. One of the requirements of the method is to minimize the size of the training data. The learning algorithm incorporates morphological language parameters, de ned by the user, that constrains the space of the induction procedure. The result is a set of morphological rules that can be inspected by the user and further compiled into a morphological analyzer (or generator).