Low-resource Morphological Generation with Neural Sequence-to-Sequence Models

Explore morphological generation techniques for low-resource languages in this 47-minute conference talk by Katharina Kann from the Center for Language & Speech Processing at JHU. Delve into neural sequence-to-sequence models for morphological inflection and reinflection tasks, with a focus on character-based approaches. Learn strategies to overcome the challenges of limited training data in morphologically rich languages, including multi-task learning, cross-lingual transfer learning, and semi-supervised learning methods. Gain insights from Kann's award-winning research in the SIGMORPHON shared tasks on morphological reinflection. Discover how these techniques can improve NLP capabilities for languages beyond English, addressing the growing importance of accurate morphology handling in diverse linguistic contexts.

Low-resource Morphological Generation with Neural Sequence-to-Sequence Models -- Katharina Kann 2017