In-Domain African Languages Translation Using LLMs and Multi-armed Bandits

TL;DR

This paper explores using bandit algorithms to select the best neural machine translation model for African languages in domain-specific tasks, especially when data is scarce or unavailable, improving translation quality.

Contribution

It introduces a bandit-based model selection framework tailored for low-resource African language translation, addressing domain adaptation without extensive fine-tuning.

Findings

Bandit algorithms effectively select optimal models for African language translation.

The approach performs well with limited or no target domain data.

Results demonstrate robustness across multiple languages and domains.

Abstract

Neural Machine Translation (NMT) systems face significant challenges when working with low-resource languages, particularly in domain adaptation tasks. These difficulties arise due to limited training data and suboptimal model generalization, As a result, selecting an optimal model for translation is crucial for achieving strong performance on in-domain data, particularly in scenarios where fine-tuning is not feasible or practical. In this paper, we investigate strategies for selecting the most suitable NMT model for a given domain using bandit-based algorithms, including Upper Confidence Bound, Linear UCB, Neural Linear Bandit, and Thompson Sampling. Our method effectively addresses the resource constraints by facilitating optimal model selection with high confidence. We evaluate the approach across three African languages and domains, demonstrating its robustness and effectiveness in both scenarios where target data is available and where it is absent.