Analyzing and Improving Cross-lingual Knowledge Transfer for Machine Translation

TL;DR

This paper investigates cross-lingual knowledge transfer in neural machine translation, analyzing factors affecting transfer effectiveness and proposing methods to enhance robustness and generalization for low-resource languages.

Contribution

It provides new insights into how language similarity, data diversity, and training strategies impact multilingual model performance and introduces techniques to improve transfer and reduce off-target errors.

Findings

Increasing translation coverage enhances generalization.

Language similarity influences transfer effectiveness.

Auxiliary supervision strengthens low-resource translation.

Abstract

Multilingual machine translation systems aim to make knowledge accessible across languages, yet learning effective cross-lingual representations remains challenging. These challenges are especially pronounced for low-resource languages, where limited parallel data constrains generalization and transfer. Understanding how multilingual models share knowledge across languages requires examining the interaction between representations, data availability, and training strategies. In this thesis, we study cross-lingual knowledge transfer in neural models and develop methods to improve robustness and generalization in multilingual settings, using machine translation as a central testbed. We analyze how similarity between languages influences transfer, how retrieval and auxiliary supervision can strengthen low-resource translation, and how fine-tuning on parallel data can introduce unintended trade-offs in large language models. We further examine the role of language diversity during training and show that increasing translation coverage improves generalization and reduces off-target behavior. Together, this work highlights how modeling choices and data composition shape multilingual learning and offers insights toward more inclusive and resilient multilingual NLP systems.