DEVELOPING AN AUTOMATED AUDIO TRANSLATION SYSTEM FOR EWE TO ENGLISH
Faculty
Year of Publication
Keyword
upload
Publication Type
Abstract
This paper presents the development of an Automated Audio Translation System (AATS) specifically designed for translating spoken Ewe, a Western Nigerian language predominantly spoken within the Yoruba people into English. The project addresses the growing need for effective communication tools in multilingual contexts, particularly in regions where Ewe is widely spoken. The proposed system leverages advanced machine learning techniques, including automatic speech recognition (ASR), natural language processing (NLP), and neural machine translation (NMT) to facilitate real-time audio translation.
The methodology involves the collection of a diverse corpus of Ewe audio recordings paired with their English translations, which serves as the training dataset for the ASR and NMT components. We employ deep learning architectures, such as recurrent neural networks (RNNs) and transformer models, to enhance the accuracy and fluency of the translation output. Evaluation metrics, including Word Error Rate (WER) and BLEU scores, are utilized to assess the performance of the
system against baseline models.
Preliminary results indicate that the AATS achieves a significant improvement in translation
accuracy compared to traditional translation methods. This research not only contributes to the field of computational linguistics but also aims to promote cultural exchange and accessibility by
bridging language barriers
The methodology involves the collection of a diverse corpus of Ewe audio recordings paired with their English translations, which serves as the training dataset for the ASR and NMT components. We employ deep learning architectures, such as recurrent neural networks (RNNs) and transformer models, to enhance the accuracy and fluency of the translation output. Evaluation metrics, including Word Error Rate (WER) and BLEU scores, are utilized to assess the performance of the
system against baseline models.
Preliminary results indicate that the AATS achieves a significant improvement in translation
accuracy compared to traditional translation methods. This research not only contributes to the field of computational linguistics but also aims to promote cultural exchange and accessibility by
bridging language barriers
Supervisor(s)
co-supervisor