Incremental Learning of Affix Segmentation

Proceedings of COLING 2012: Technical Papers, pages 1901–1914, COLING 2012, Mumbai, December 2012. Incremental Learning of Affix Segmentation Wondwossen Mulugeta1, Michael Gasser2, Baye Yimam1 (1) ADDIS ABABA UNIVERSITY, Addis Ababa, Ethiopia (2) INDIANA UNIVERSITY, Bloomington, USA wondisho@aau.edu.et, gasser@cs.indiana.edu, bayemekonnen@gmail.com ABSTRACT This paper presents a supervised machine learning approach to incrementally learn and segment affixes using generic background knowledge. We used Prolog script to split affixes from the Amharic word for further morphological analysis. Amharic, a Semitic language, has very complex inflectional and derivational verb morphology, with many possible prefixes and suffixes which are used to show various grammatical features. Further segmentation of the affixes into valid morphemes is a challenge addressed in this paper. The paper demonstrates how incremental and easy-to-complex examples can be used to learn such language constructs. The experiment revealed that affixes could be further segmented into valid prefixes and suffixes using a generic and robust string manipulation script by the help of an intelligent teacher who presents examples in incremental order of complexity allowing the system to gradually build its knowledge. The system is able to do the segmentation with 0.94 Precision and 0.97 Recall rates. KEYWORDS: Amharic, Morphology, Segmentation, Incremental Learning, ILP, Machine Learning 1901 1 Introduction Amharic is a Semitic language, related to Hebrew, Arabic, and Syriac. Next to Arabic, it is the second most spoken Semitic language with around 27 million speakers (Sieber, 2005; Gasser, 2011). As the working language of the Ethiopian Federal Government and some regional governments in Ethiopia1, most documents in the country are produced in Amharic. There is also an enormous production of electronic and online accessible Amharic documents. One of the fundamental computational tasks for a language is analysis of its morphology, where the goal is to derive the root and grammatical properties of a word based on its internal structure. Morphological analysis, especially for complex languages like Amharic, is vital for development and application of many practical natural language processing systems such as machine-readable dictionaries, machine translation, information retrieval, spell-checkers, and speech recognition. While various approaches have been used for other languages, Amharic morphology has so far been attempted using only rule-based methods. In our previous work, we have tried to apply a machine learning approach to learn morphological rules. In the experiment, we were able to learn various affixes attached to the stem and analyze the internal stem structure of the verb which is one crucial task in Semitic morphology. The major limitation of the work concerns words made up of the stem and more than one adjacent prefix or suffix; in those cases the system fails to segment the affixes. This work presents the continuation of our previous system and attempts to further segment the affixes into valid prefixes and suffixes using generic and incremental learning without any initial knowledge of the prefixes and suffixes of the language. 2 Amharic Verb Morphology and Affixation Amharic, with all its complex word formation nature, has been more or less thoroughly studied by linguists (Sieber, 2005; Dawkins, 1960; …

Content truncated to 500 words