Efficient Multilingual and Domain Adaptation of Language Models Under Resource Constraints
MCML Authors
Alexandra Chronopoulou
Dr.
* Former Member
Abstract
Alexandra Chronopoulou
Dr.
* Former Member
Abstract
This dissertation develops methods to improve natural language processing (NLP) systems for low-resource languages and diverse domains. For machine translation, it explores bilingual language models, static embeddings, and multilingual systems with adapters, achieving robust performance in low-resource settings. To enhance domain adaptation, it introduces hierarchical tree structures and efficient adapters, enabling better generalization and robustness to domain shifts. These approaches address data disparities and domain variability, advancing adaptable and efficient NLP systems. (Shortened).
BibTeXKey: Chr24