A 3-Steps Algorithm for Morphological Disambiguation Using Untagged Corpora

This article presents a three steps algorithm for morphological disambiguation between the definite article and the personal pronoun in French language. Tested accuracy in a large untagged corpora exceeds 98% with less than 1% of error. Our method has been also experimented on unlabeled Greek corpora and the results prove the system’s portability to other languages with similar structure. Not any prior knowledge is available. The rule-based procedure is robust and selfcorrecting. It can also be used as a shallow parser for verbal and nominal groups identification. The last step of the algorithm consists on the creation of a dictionary with classification of the entries in two grammatical categories : nominal and verbal.