Computer Speech and Language 82 (2023) 101524 Available online 12 May 2023 0885-2308/© 2023 Elsevier Ltd. All rights reserved. Contents lists available at ScienceDirect Computer Speech & Language journal homepage: www.elsevier.com/locate/csl English–Assamese neural machine translation using prior alignment and pre-trained language model Sahinur Rahman Laskar a , Bishwaraj Paul a , Pankaj Dadure b , Riyanka Manna c , Partha Pakray a,∗ , Sivaji Bandyopadhyay a a Department of Computer Science and Engineering, National Institute of Technology, Silchar, 788010, Assam, India b School of Computer Science, University of Petroleum and Energy Studies, Dehradun, 248007, Uttarakhand, India c Department of Computer Science and Engineering, Jadavpur University, Kolkata, 700032, West Bengal, India ARTICLE INFO Keywords: Low-resource NMT English–Assamese Alignment Language model ABSTRACT In a multilingual country like India, automatic natural language translation plays a key role in building a community with different linguistic people. Many researchers have explored and improved the translation process for high-resource languages such as English, German, etc., and achieved state-of-the-art results. However, the unavailability of adequate data is the prime obstacle to automatic natural language translation of low-resource north-eastern Indian languages such as Mizo, Khasi, and Assamese. Though the recent past has witnessed a deluge in several automatic natural language translation systems for low-resource languages, the low values of their evaluation measures indicate the scope for improvement. In the recent past, the neural machine translation approach has significantly improved translation quality, and the credit goes to the availability of a huge amount of data. Subsequently, the neural machine translation approach for low-resource language is underrepresented due to the unavailability of adequate data. In this work, we have considered a low-resource English–Assamese pair using the transformer-based neural machine translation, which leverages the use of prior alignment and a pre-trained language model. To extract alignment information from the source–target sentences, we have used the pre-trained multilingual contextual embeddings-based alignment technique. Also, the transformer-based language model is built using monolingual target sentences. With the use of both prior alignment and a pre-trained language model, the transformer-based neural machine translation model shows improvement, and we have achieved state-of-the-art results for the English-to-Assamese and Assamese-to-English translation, respectively. 1. Introduction Machine translation (MT) is a popular task of natural language processing (NLP), and it has come into the limelight in the last few decades. MT aims to perform the automatic translation from one natural language to another. The MT approaches are categorized into two broad categories: rule-based and corpus-based. The rule-based approach uses the grammatical and linguistic rules for particular language pairs to generate target translations (Saini and Sahula, 2015). However, the corpus-based approaches, namely, example-based machine translation (Somers, 1999), statistical machine translation (SMT) (Koehn et al., 2003) and neural machine translation (NMT) (Bahdanau et al., 2015; Luong et al., 2015) eliminate the need for construction of linguistic rules and reliance on linguistic experts. ∗ Corresponding author. E-mail address: partha@cse.nits.ac.in (P. Pakray). https://doi.org/10.1016/j.csl.2023.101524 Received 5 July 2022; Received in revised form 6 March 2023; Accepted 3 May 2023