Towards Multilinguality: Overcoming Language and Script Barriers in Natural Language Processing
MCML Authors
Abstract
Abstract
This dissertation advances multilingual NLP by developing methods that improve language model coverage, cross-lingual transfer, and support for underrepresented languages and scripts. It introduces multilingual concept graphs, more efficient pretraining strategies, and transliteration-based learning objectives that enhance alignment across languages, making NLP technologies more inclusive and effective in highly multilingual settings. (Shortened.)
BibTeXKey: Liu26