Stanford CS224N: NLP with Deep Learning | Winter 2019 | Lecture 15 – Natural Language Generation

For more information about Stanford’s Artificial Intelligence professional and graduate programs, visit: https://stanford.io/3Cfhyya

Professor Christopher Manning & PhD Candidate Abigail See, Stanford University
http://onlinehub.stanford.edu/

Professor Christopher Manning
Thomas M. Siebel Professor in Machine Learning, Professor of Linguistics and of Computer Science
Director, Stanford Artificial Intelligence Laboratory (SAIL)

To follow along with the course schedule and syllabus, visit: http://web.stanford.edu/class/....cs224n/index.html#sc

0:00 Introduction
0:27 Announcements
1:09 Overview
2:47 Natural Language Generation (NLG)
5:00 Recap: training a (conditional) RNN-LM
6:21 Recap: decoding algorithms
6:47 Recap: greedy decoding
7:32 Recap: beam search decoding
8:57 Aside: Do the hosts in Westworld use beam search?
10:07 What's the effect of changing beam size k?
13:16 Effect of beam size in chitchat dialogue
15:58 Sampling-based decoding
18:22 Softmax temperature
21:03 Decoding algorithms: in summary
22:55 Summarization: task definition
27:04 Summarization: two main strategies
28:20 Pre-neural summarization
31:00 Summarization evaluation: ROUGE
35:20 Neural summarization (2015-present)
38:53 Neural summarization: copy mechanisms
42:58 Neural summarization: better content selection
43:47 Bottom-up summarization
45:46 Neural summarization via Reinforcement Learning
49:36 Pre- and post-neural dialogue
50:56 Seq2seq-based dialogue
52:36 Irrelevant response problem
54:19 Genericness / boring response problem
56:38 Repetition problem
59:12 Storytelling
59:35 Generating a story from an image