Natural Language Generation

This lecture covers the fundamentals of natural language generation (NLG), a sub-field of natural language processing focused on building systems that automatically produce coherent and useful written or spoken text for human consumption. Topics include formalizing NLG, decoding from NLG models, evaluating NLG systems, and various decoding methods such as greedy decoding, beam search, and sampling methods. The lecture also delves into content overlap metrics, model-based metrics, and human evaluations to assess the quality of generated text.