NLIP_Lab-IITH Low-Resource MT System for WMT24 Indic MT Shared Task

TL;DR

This paper presents a low-resource Indic language translation system for WMT24, utilizing fine-tuning of pre-trained models with alignment augmentation, achieving competitive scores across multiple language pairs.

Contribution

It introduces a language-specific fine-tuning approach with alignment augmentation for low-resource Indic languages in machine translation.

Findings

Achieved high chrF2 scores for four Indic language pairs.

Explored multilingual training with various strategies.

Provided open-source code and models for reproducibility.

Abstract

In this paper, we describe our system for the WMT 24 shared task of Low-Resource Indic Language Translation. We consider eng $\leftrightarrow$ {as, kha, lus, mni} as participating language pairs. In this shared task, we explore the finetuning of a pre-trained model motivated by the pre-trained objective of aligning embeddings closer by alignment augmentation \cite{lin-etal-2020-pre} for 22 scheduled Indian languages. Our primary system is based on language-specific finetuning on a pre-trained model. We achieve chrF2 scores of 50.6, 42.3, 54.9, and 66.3 on the official public test set for eng$\rightarrow$as, eng$\rightarrow$kha, eng$\rightarrow$lus, eng$\rightarrow$mni respectively. We also explore multilingual training with/without language grouping and layer-freezing. Our code, models, and generated translations are available here: https://github.com/pramitsahoo/WMT2024-LRILT.