English-Amharic Document Translation Using Hybrid Approach
No Thumbnail Available
Date
2017-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Machine Translation (MT) is a sub field of Natural Language Processing (NLP) that
investigates the use of computer software to translate text or speech from one natural
language to another. Recently, in Ethiopia English documents are translated to Amharic
using human translators, because of the absence of Automated Translation System. Due
to this, the process of document translation is so expensive, challenging, unsecured and
time consuming. To solve those problems statistical based approaches were proposed to
translate English to Amharic. However, the approaches have accuracy and understandability
issues. In order to solve these problems we propose a Hybrid approach machine
translation (HMT) system that combines Statistical and Rule Based Machine Translation
approaches.
In this case using the hybrid approach achieved better accuracy and found to be advanced
for English - Amharic machine translation system over SMT approach. The proposed
hybrid approach achieved 15% and 20% accuracy improvement for simple and complex
sentences over statistical machine translation approach. This study also identified preprocessing
the inputs of SMT is more suitable to improve accuracy for complex sentences
while post-processing the outputs of SMT is more suitable for simple sentences.
Description
Keywords
Machine Translation, SMT, RBMT, HMT