An Exploration of Data Augmentation Techniques for Improving English to Tigrinya Translation

TL;DR

This paper investigates various back-translation methods to enhance English to Tigrinya translation, demonstrating that pivoting through a related high-resource language significantly improves performance in low-resource scenarios.

Contribution

The study provides a detailed analysis of back-translation techniques for Tigrinya, highlighting the effectiveness of pivoting through related languages in low-resource translation tasks.

Findings

Pivot-based back-translation yields the best improvements.

Synthetic data significantly boosts translation quality.

Low-resource conditions benefit most from related language pivoting.

Abstract

It has been shown that the performance of neural machine translation (NMT) drops starkly in low-resource conditions, often requiring large amounts of auxiliary data to achieve competitive results. An effective method of generating auxiliary data is back-translation of target language sentences. In this work, we present a case study of Tigrinya where we investigate several back-translation methods to generate synthetic source sentences. We find that in low-resource conditions, back-translation by pivoting through a higher-resource language related to the target language proves most effective resulting in substantial improvements over baselines.