LING 575: Care and feeding of neural networks for NLP
(subtitle: How to Train Your Dragon)
Instructor: Jeremy Kahn
27 March-29 May 2017 (inclusive).
First day of instruction: March 27
This course offers an introductory exploration of neural networks for natural language processing. Neural networks are enjoying recent extraordinary success in many domains; in this course, we explore these approaches to classifiers, embeddings, and sequence-modeling applications. Students will explore (and implement!) the cross-product of network shapes, training strategy, training corpora, and evaluation targets. Neural networks -- like many objective-driven approaches -- are notoriously opaque to inspection, so this course will have a special focus on adaptation and domain-adaptable models and the reuse or re-training of existing models.