Machine Translation System for Amharic Text to Ethiopian sign language
No Thumbnail Available
Date
2011-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Machine Translation (MT) system development from one language to another is one of the areas in human computer interaction researches for the last half century. In this thesis the first Amharic to Ethiopian Sign Language (ESL) MT system was designed and implemented for limited vocabulary. The MT system function is to change or translate a given Amharic text words to their equivalent representation in ESL. To achieve this goal the translation engine should have subsystems to analyze source language (Amharic) and synthesize target language (ESL). Due to close relation of source and target language, in addition to the system is designed for limited vocabulary, Rule Based Machine Translation (RBMT) approach is deployed in this research. Since there was no enough information on grammar of ESL, this thesis scope is limited to morphological level translation; i.e. direct RBMT architecture was followed. While designing Amharic language morphological analyzer subsystem, it was found that Amharic alphabets (fidel) rearrangement based on their phonetic behavior gave a much needed pattern to apply expansion rules. In this thesis, a new arrangement of fidel was proposed to be used in computer based researches on Amharic language which involve morphology or phonology. Amharic language is morphologically rich language, so it‟s not easy to prepare bilingual dictionary with all possible word derivations. To overcome this problem morphological expansion method was used. In this particular case 40 rules were applied on 20 verbs and 3 rules applied on 10 nouns to get a dictionary with entries more than 800. Many Sign Language(SL) translation researches, nowadays use universal transcription methods; HamNoSys and SiGML. This thesis also applied these transcription tools for the first time on ESL. For a randomly selected fifty words and all alphabets in ESL, the transcription was done.
The system was designed to map a given Amharic word to ESL equivalent words using
ix
morphological analysis and synthesis. If the word is not in the bilingual dictionary, it‟ll finger spell rather than ignoring it. And performance of the system was found to be more than 80%.
keywords- Sign Language, ESL, Amharic morphology, machine translation, Deaf.
Description
Keywords
Sign Language, ESL, Amharic morphology, machine translation, Deaf