Gex'ez-English Bi-Directional Neural Machine Translation Using Transformer | IEEE Conference Publication | IEEE Xplore

Gex'ez-English Bi-Directional Neural Machine Translation Using Transformer


Abstract:

Machine translation is the technique of translating texts from one language to another without human intervention using artificial intelligence. Neural Machine Translatio...Show More

Abstract:

Machine translation is the technique of translating texts from one language to another without human intervention using artificial intelligence. Neural Machine Translation (NMT) is a method of machine translation that employs a large artificial neural network such as Transformers to forecast the possibility of a set of words, frequently in the form of complete sentences. There are several old Ge'ez scripts both in Ethiopia and elsewhere that require translation. Students and researchers are currently eager in learning about and being involved in the study of manuscripts written in Ge'ez. To increase Ge'ez users and solve the problems within the endangered Ge'ez language, it is important to develop Neural Machine Translation between English and Ge'ez languages. Therefore, the purpose of this paper is to propose the Ge'ez to English Bi-directional Neural Machine Translation using a deep learning approach. We carried out our study by investigating the capabilities of deep learning algorithms of the Transformer models. We have gathered 16,569 parallel corpus data from various sources and separated them into training and testing sets to accomplish our goal. 80% of the whole dataset was used for training, while the remaining 20% was used for testing. We applied preprocessing techniques like cleaning, normalization, tokenization, and padding. BLEU score evaluation metrics have been used. The transformer model achieved 27.19% and 29.39% BLEU scores from English to Ge'ez and Ge'ez to English translation respectively. The major challenge of this paper was the scarcity of enough datasets to conduct an extensive experiment and get better results.
Date of Conference: 26-28 October 2023
Date Added to IEEE Xplore: 06 November 2023
ISBN Information:
Conference Location: Bahir Dar, Ethiopia

References

References is not available for this document.