Triangular Transfer: Freezing the Pivot for Triangular Machine Translation

TL;DR

This paper introduces a transfer learning method for low-resource triangular machine translation that freezes the pivot language model parameters to better leverage auxiliary data, improving translation quality.

Contribution

It proposes freezing the pivot language model during training to enhance transfer learning in triangular machine translation tasks.

Findings

Outperforms previous methods in low-resource settings

Effective utilization of auxiliary data improves translation quality

Freezing pivot model parameters stabilizes training

Abstract

Triangular machine translation is a special case of low-resource machine translation where the language pair of interest has limited parallel data, but both languages have abundant parallel data with a pivot language. Naturally, the key to triangular machine translation is the successful exploitation of such auxiliary data. In this work, we propose a transfer-learning-based approach that utilizes all types of auxiliary data. As we train auxiliary source-pivot and pivot-target translation models, we initialize some parameters of the pivot side with a pre-trained language model and freeze them to encourage both translation models to work in the same pivot language space, so that they can be smoothly transferred to the source-target translation model. Experiments show that our approach can outperform previous ones.