Finite-state transducer for Amazigh verbal morphology

摘要

With the aim of promoting Amazigh language through human language technology, the Royal Institute of Amazigh Culture is investing a great deal of effort in producing necessary linguistic resources and tools. In this context, this article gives an overview of an original Amazigh verbal morphological processor based on Xerox's finite-state transducer platform. This processor consists of two main components: a lexicon including 3,637 attested verbs, and a set of rules covering the morphotactic and morphophonological phenomena observed in standard Amazigh texts. Its reliance on finite-state technology has ensured the implementation of an efficient bidirectional system for analysis and generation, achieving a rate of 91 of correct results in the analysis of standardized verbs. This Amazigh verbal analyzer/generator will significantly improve the performance of many natural language processing applications such as the Amazigh information retrieval system (Ataa Allah, and Boulaknadel, 2010a) and Amazigh concordancer (Ataa Allah, and Boulaknadel, 2010b) and will also certainly contribute to the implementation of other applications such as spellchecking, machine translation, and computer-aided language learning.