Agglutinative Morphologies
This paper presents a method for analysing word forms in highly-inflectional and agglutinative natural languages. The method described is language independent and can be run either automatically or as a semi-supervised process. The wider context of this research in providing authoring tools for developing world vernacular languages is discussed. Other work in the field of automatic morpheme analysis is reviewed and its relevance to this method and context evaluated.

Morpheme identification is made by means of establishing the relationship of each hypothesised morpheme with its corresponding stem lemmata. A list of discovered morphemes is generated together with a list of their stems. Inflection paradigms are identified where stems share the same morpheme structures. The process is described in detailed and three example analyses are shown drawn from disparate language groups. The language independent nature of the process is demonstrated. The benefits and limitations of the method are discussed and areas for further work identified.

Bartholome processing a Latin word list.

Automatic Morphology Analysis v0.1

Cite this work:
Masters thesis (Riding2008)
Riding, J. D.
A Relational Method for the Automatic Analysis of Highly-Inflectional Agglutinative Morphologies
Oxford Brookes University, Computing & Mathematical Sciences, Pattern Recognition Group., 2007

Back to category overview   Back to news overview