To Sentences and Beyond! Paving the Way for Context-Aware Machine Translation

Explore the limitations of sentence-level machine translation and the path towards context-aware translation in this insightful 41-minute presentation by Rachel Wicks from the Center for Language & Speech Processing at Johns Hopkins University. Delve into the challenges posed by sentence segmentation in the translation pipeline and understand why perfect sentence-level machine translation is an unattainable goal. Examine the importance of inter-sentential context for accurate translation of grammatical features. Discover the major obstacle hindering true context-aware machine translation - the severe lack of suitable data. Learn about recent advancements in the field, including a new evaluation dataset specifically designed to address context-dependent discourse phenomena in translation. Gain insights into innovative methods for reconstructing documents from large-scale sentence-level bitext, which can enhance performance when translating context-dependent elements.

To Sentences and Beyond! Paving the way for Context-Aware Machine Translation