International Journal of Computing and Digital Systems ISSN (2210-142X) Int. J. Com. Dig. Sys. 15, No.1 (Apr-24) http://dx.doi.org/10.12785/ijcds/1501107 A Survey on the MT Methods for Indian Languages: MT Challenges, Availability, and Production of Parallel Corpora, Government Policies and Research Directions Sudeshna Sani 1 , Samudra Vijaya 2 and Suryakanth V Gangashetty 3 1,2,3 Department of CSE, Koneru Lakshmaiah Education Foundation, Guntur, Andhra Pradesh, India Received 13 Dec. 2023, Revised 14 Mar. 2024, Accepted 18 Mar. 2024, Published 1 Apr. 2024 Abstract: Since 1991, machine translation has been a prominent research area in India, with IIT Kanpur pioneering the original work which has since been expanded to several universities. Only 10 percent of India’s 1.3 billion inhabitants can read, write, and speak English with varying degrees of competence, which makes machine translation crucial in overcoming the linguistic barrier to the internet. The Indian market for commercial products and events is greatly influenced by local languages, making the development and translation of region-based content an essential research topic nowadays. However, Indic-to-Indic language direct translation has faced several challenges and is still going through the experimental phase. Several government-sponsored projects are being undertaken in this regard. Still, there are limited sentence-aligned parallel bi-text resources available for the majority of Indian language pairs. This paper presents a detailed survey of the current trends of research on machine translation between Indian languages, along with their challenges over time. It also presents a timeline of recent research conducted and key findings of past surveys conducted over a decade. Under a single canopy, this paper provides sources of data, the progress made in developing datasets for low-resource Indian languages, various models of translation, encouragement from Indian Govt., and finally, new research directions. Keywords: Machine Translation, RBMT, SMT, NMT, Low-Resource Indian languages,BLUE, METEOR, AI4Bharat, Bhashini 1. Introduction Machine Translation (MT) is a method of translating one written human language automatically in to another language, while maintaining the significance of the source text and generating fluent and proper text in the target language. MT has been developed as a subfield of Artificial Intelligence (AI) and is a part of computational linguis- tics and language engineering. MT techniques are further improved by utilizing concepts and methods from various fields such as statistics, computer science, AI, translation theory, and linguistics [1]. Figure 1 shows the basic structure of an MT system. Machine Translation (MT) research in Indian languages is relatively less developed as compared to other interna- tional languages such as English, Chinese, and Spanish. This is primarily due to the complexity and diversity of Indian languages, which makes MT a challenging task. Additionally, Indian languages have low resource availabil- ity, lack of parallel corpora, and limited research funding. However, in recent years, a growing MT research interest Figure 1. Diagram of a Basic MT System for Indian languages is observed, with several initiatives and collaborations between academia, industry, and government. Various research projects are underway to advance MT systems for Indian languages, and efforts are being made to increase the availability and quality of parallel corpora for Indian languages. Despite the challenges, MT research in Indian languages has great potential in the current global market scenario. India is a distinct country with more than 1.3 billion residents, and a growing economy with a huge demand for localization of content in regional languages. Indian languages are typically classified into five major language families [2] [3]: E-mail address: sudeshnasani@gmail.com, samudravijaya@kluniversity.in, svg@kluniversity.in https:// journal.uob.edu.bh/