Assessing Neural Machine Translation for English-Urdu with LSTM, GRU, And Transformer


Master of Science in Data Science


Department of Computer Science

Faculty/ School

School of Mathematics and Computer Science (SMCS)

Date of Submission

Fall 2023


Dr. Sajjad Haider, Professor, Department of Computer Science, Institute of Business Administration, Karachi


This report presents a study on the performance of various Neural Machine Translation (NMT) models for the English-Urdu pair, a relatively unexplored domain in machine translation research. The primary objective of this study is to identify the most effective NMT model for translating between these two languages.

The research focuses on experimenting with different NMT models, including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Transformer models, in different configurations to determine their efficacy in capturing the nuances of English and Urdu. The study also incorporated techniques like SentencePiece Tokenization and FastText Embedding to enhance the models' understanding of complex language structures.

Two distinct datasets were utilized for training and testing: a Parallel corpus from Kaggle, containing basic vocabulary sentence pairs, and the more extensive CVIT PIB dataset, offering a broader range of sentence structures and lengths. The models' performances were evaluated using the BLEU score metric.

The findings reveal significant insights into the capabilities and limitations of each NMT model in handling the English-Urdu translation. While some models showed promising results in certain aspects, the study highlights the need for more extensive datasets and deeper model architectures to achieve higher accuracy and fluency in translations. This report contributes to the field of NMT by providing a detailed analysis of English-Urdu translation, paving the way for future research and development in this area.

Document Type

Restricted Access

Submission Type

Research Project

This document is currently not available here.

The full text of this document is only accessible to authorized users.