Indonesian Journal of Electrical Engineering and Computer Science Vol. 27, No. 2, August 2022, pp. 1083~1090 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v27.i2.pp1083-1090  1083 Journal homepage: http://ijeecs.iaescore.com Hindi to English transliteration using multilayer gated recurrent units Mohd Zeeshan Ansari 1 , Tanvir Ahmad 1 , Mirza Mohd Sufyan Beg 2 , Faiyaz Ahmad 1 1 Department of Computer Engineering, Faculty of Engineering and Technology, Jamia Millia Islamia University, New Delhi, India 2 Department of Computer Engineering, Aligarh Muslim University, Aligarh, India Article Info ABSTRACT Article history: Received Sep 23, 2021 Revised May 25, 2022 Accepted Jun 10, 2022 Transliteration is the task of translating text from source script to target script provided that the language of the text remains the same. In this work, we perform transliteration on less explored Devanagari to Roman Hindi transliteration and its back transliteration. The neural transliteration model in this work is based on a sequence-to-sequence neural network that is composed of two major components, an encoder that transforms source language words into a meaningful representation and the decoder that is responsible for decoding the target language words. We utilize gated recurrent units (GRU) to design the multilayer encoder and decoder network. Among the several models, the multilayer model shows the best performance in terms of coupon equivalent rate (CER) and word error rate (WER). The method generates quite satisfactory predictions in Hindi- English bilingual machine transliteration with WER of 64.8% and CER of 20.1% which is a significant improvement over existing methods. Keywords: Encoder decoder Gated recurrent units Sequence to sequence model Transliteration This is an open access article under the CC BY-SA license. Corresponding Author: Faiyaz Ahmad Department of Computer Engineering, Faculty of Engineering and Technology Jamia Millia Islamia University Maulana Mohammad Ali Jauhar Marg, New Delhi, 110025, India Email: fahmad1@jmi.ac.in 1. INTRODUCTION The transliteration from a source to a target script is defined as writing the text using the letters of the target language provided that the language of the text does not change. Moreover, it preserves the pronunciation of a word while transforming it from a source script to a target script [1]. Transliteration from different languages to English is useful in bilingual knowledge extraction tasks including information retrieval, named entity recognition and automatic bilingual dictionary compilation. [2]-[4]. The out of vocabulary (OOV) words like names, and acronyms. In cross-lingual tasks are significantly transcribed into the base document language, provided that the source and target do not share the alphabet. The named entity transliteration plays a significant role in cross-language tasks, apparently, during document translation from source to target language, named entities are transliterated. Transliteration being the subtask of translation as it transforms one language script into corresponding similar phonetic characters of the target alphabet poses several challenges due to differences in syntax, morphology, and semantics between the source script and the target script language. Hindi to English transliteration or vice versa, pose dramatic challenges due to the morphologically rich nature of Hindi. For example, a Hindi word चाभी when transliterated into English has multiple transliterations of chabhi, chaabhi, chaabhee, chaabhie. Perhaps, the back transliteration is even more challenging as several words transliterate into a single target word. In this work, we employ the neural framework for transliteration, basically the sequence-to-sequence modelling based on recurrent neural