DEVELOPING A MORPHOLOGICAL ANALYZER AND GENERATOR FOR AWNGI VERB USING FINITE STATE TRANSDUCER

Abstract

Language is a powerful medium that coordinates day-to-day activities between an entity. The means of interaction between a computer and language structure is studied by natural language processing. Natural language processing has emerged as a means of increasing computers’ capability to understand natural languages. Awngi is one of the natural languages, grouped under the central Cushitic language family and spoken by more the 2.5 million people as native speakers settled in a different part of the Amhara regional state. It is a zona language that started as a medium of instruction in 1989 E.C. and uses the Geez script system. It is the one of under-resourced languages in Ethiopia. This morphological study is the first attempt that is used as a contribution to high-level NLP applications. The study is an implementation of the morphological analyzer and generation based on a finite-state transducer approach. The lexicon formalism is used to design the morphological analyzer and generation of the Awngi verb form because the Awngi language is morphologically very productive in terms of gender, person, number, and tenses. The system was developed using the foma programing language as an implementation toolkit. The foma programing language and the Ubuntu 22.04 Linux-based operating system are used for experimentation. For the experimental thirty-five distinct linguistic rules were developed to form those different word forms. The experimentshowsthat 76956 words were generated by the system out of this 93.6% of words are correctly generated and analyzed by the system and 6.4% of words were wrongly generated. The factors that the wrongly generated words were the problem of the epenthetic sound vowel እ/ɨ/ breaking up consonant clusters of words and the glottal approximant /h/, which is loaned from Amharic appears irregularly in the Awngi language word formation process. So, determining the correct phonological pattern of the glottal approximate /h/ and the alternative techniques for the epenthetic vowel እ/ɨ/ management is left for future work