Learning rules for morphological analysis and synthesis of Macedonian nouns

Abstract

This paper presents a machine learning approach to morphological analysis and synthesis of Macedonian nouns. For training and testing we used the nouns originating from Orwell’s “1984”. The paper presents experimental results of using the learned rules in the process of analysis, and in the process of noun formation. Training was performed with the whole set of Macedonian nouns from “1984” and tested by 10-fold cross-validation. All the potential nouns forms generated by the learning rules were compared with 275000 Macedonian noun forms. The accuracy of 92-97% is encouraging to apply the same approach to all categories of Macedonian words.