Litcius/Paper detail

Performance Evaluation of Popular Deep Neural Networks for Neural Machine Translation

Muhammad Naeem, Abu Bakar Siddique, Raja Hashim Ali, Usama Arshad, Zain Ul Abideen, Talha Ali Khan, Muhammad Huzaifa Shah, Ali Zeeshan Ijaz, Nisar Ali

202330 citationsDOI

Abstract

The field of Neural Machine Translation (NMT) has shown impressive performance for quick and easy communication in various languages spoken all over the world. NMT helps us by improving communication between different languages. For this purpose, different sequential models are used such as Recurrent Neural Network (RNN), Long Short Term Memory (LSTM), and Gated Recurrent Units (GRU). Analysis among these different models are important for making language translation better and choose the best model for the right job. This research investigates the performance of these models on two distinct language datasets, English-to-German and English-to-Urdu. Based on accuracy metrics, the findings reveal that GRU having test accuracy (88.22% ) outperforms RNN (87.21% ), and LSTM (85.70% )demonstrating the highest translation accuracy, followed by RNN and LSTM exhibiting comparatively lower accuracy levels.

Topics & Concepts

Computer scienceMachine translationRecurrent neural networkArtificial intelligenceArtificial neural networkNatural language processingLanguage modelField (mathematics)GermanTranslation (biology)Machine learningLong short term memorySpeech recognitionGeneMessenger RNAPure mathematicsHistoryBiochemistryMathematicsChemistryArchaeologyNatural Language Processing TechniquesTopic ModelingMachine Learning and Data Classification
Performance Evaluation of Popular Deep Neural Networks for Neural Machine Translation | Litcius