A. Kadriu (Macedonia)
Machine learning, finite state automata, two-levelmorphology, Albanian noun declension
This paper describes an application of a machine learning algorithm to define a two-level formalism for the declension of the Albanian nouns. The algorithm uses minimal acyclic finite automata. The input for this system is a set of pairs base_word–inflected_word. The algorithm consists of two parts: segmentation of the inflected words and finding the optimal two-level rules with minimal context. Inflectional suffixes for five inflectional cases and five two-level rules are automatically obtained. The description has been implemented using the PC_KIMMO environment. We also illustrate some limitations and problems with this description.
Important Links:
Go Back
