Bi-Directional English-Afan Oromo Machine Translation Using Convolutional Neural Network
No Thumbnail Available
Date
2019-10-14
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Many languages are spoken across the world which can bring communication gaps where two
people that speak different languages cannot communicate. Usually, this communication gap is
solved by using a human interpreter. However, the use of human interpreters is expensive and
inconvenient. Many researches are being done to resolve this problem using machine translation
techniques. Machine translation is an automatic translation of a source language to a target
language. This can be speech to speech or text to text translation. In this work, a bi-directional
text based machine translation for English and Afan Oromo languages pair using convolutional
neural networks is proposed. We started our study with objective of improving the previous work
on English to Afan Oromo machine translation by making the translation bi-directional by
applying convolutional neural network on translations between these language pair. In order to
achieve our objective, we collected parallel corpus data from different sources and divided into
training and testing sets. We have used 80% of total dataset for training and 20% of total dataset
for testing. Three systems were implemented where the first system uses a word based statistical
approach that used as a baseline, while the second system with recurrent neural network
approach is used as a competitive model and lastly, the third system with convolutional neural
networks for the bi-directional translation between Afan Oromo and English languages.
After training and testing these systems on corresponding training and testing datasets, the
convolutional neural network achieved 3.86 BLEU score improvement on translation from
English to Afan Oromo and 3.32 BLEU score on translation from Afan Oromo to English
translation than baseline system. Also convolutional neural network approach has shown an
improvement of 1.58 BLEU score on translation from English to Afan Oromo and 1.51 BLEU
score on translation from Afan Oromo to English translation than recurrent neural network
approach. The convolutional neural network approach is faster on training than recurrent neural
network approach.
Description
Keywords
Machine translation, English-Afan Oromo machine translation, recurrent neural network, convolutional neural network bi-directional machine translation