Example Based Amharic Text to Ethiopian Sign Language Machine Translation
No Thumbnail Available
Date
2018-06-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Ethiopian sign language (EthSL) is an independent language used by the Ethiopian Deaf
society. Sign language is a visual gestural language, used by the Deaf for everyday
communication. In most societies sign language is not known, and if known, it is not well
understood even by their own family. This in turn makes their communication to be dependent
on interpreters even in situations like medical treatment that need privacy.
Machine translation for spoken languages is a successful area of research and development.
But in case of sign language, especially in case of EthSL it is in its infancy. This is because
Sign language research is still a relatively new area when compared to research into spoken
languages.
In this study an attempt is made to design and implement an example based Amharic Text to
Ethiopian Sign Language Machine Translation system by collecting examples. The example
data is processed to produce morphologically analyzed Amharic words, word structures and
sentence structures in addition to adopting EthSL dictionary. A prototype performing
translation based on knowledge acquired from processed example data is thus developed by
involving word feature analysis, matching, alignment and recombination phases. The resulting
translation output of EthSL sign is displayed as a sequence of video clips.
The system is assessed in terms of adequacy and fluency with three groups of evaluators for
different categories of texts. The evaluation result shows that finger spellings, numbers and
words from the adopted dictionary were adequately understood with a rank of ‘All meaning’
and with ‘Good’ fluency whereas words out of the adopted dictionary were adequately
understood with a rank of ‘Most meaning’ with ‘Acceptable’ fluency. In case of sentences
‘Much meaning’ has been adequately understood with ‘Acceptable’ fluency level.
The usability of the prototype was ranked as ‘Excellent’ with a ‘Very good’ level of ease of
use. Deaf members of the test teams have been aspired with the capability of the prototype for
future use and they expressed their willingness to contribute for full deployment of the system.
Description
Keywords
Ethiopian Sign Language, Sign Language, Example Based Machine Translation