Schedule

Note: Future schedule is subject to minor change.

Assignments

• Assignment 1: Representing and classifying text hw1.tgz hw1.pdf (Due on Sep 29)

• Assignment 2: Predicting sequences hw2.tgz hw2.pdf (Due on Oct 13)

• Assignment 3: Hidden Markov models and EM hw3.tgz hw3.pdf (Due on Nov 8)

• Assignment 4: Constituent parsing hw4.tgz hw4.pdf (Due on Nov 24)

Introduction

• Week 1 (Sep 8). Overview: NLP tasks and challenges, basic ML

  • slides 1 slides 2 notes

Representation of text

• Week 2 (Sep 15). Text classification: bag-of-words, naive Bayes models, logistic regression

  • slides board recording notes

• Week 3 (Sep 22). Distributed representation: vector space models, Brown clusters, neural word embeddings

  • slides board recording notes

Predicting sequences

• Week 4 (Sep 29). Language models: n-gram LM, neural LM, perplexity

  • slides board recording notes

• Week 5 (Oct 6). Sequence labeling: log-linear models, decoding, POS tagging

  • slides recording notes

• Week 6 (Oct 13). Hidden Markov models: HMM, EM

  • slides recording J&M HMM, Collins EM

• Week 7 (Oct 20). Midterm.

Predicting trees

• Week 8 (Oct 27). Context-free parsing: PCFG, CYK, neural parser

  • slides recording Collins PCFG, Eisner Inside-Outside

• Week 9 (Nov 3). Semantic parsing: logical semantics, learning from logical forms / denotations

  • slides recording E Ch12, Liang 16

Deep learning for NLP

• Week 10 (Nov 10). Neural sequence modeling: seq2seq, attention, copy mechanism, text generation

  • slides recording D2L 9.7, 9.8, 10

• Week 11 (Nov 17). Representation learning: transformers, contextualized word embedding, pre-training and fine-tuning, autoencoders

  • slides recording

  • Illustrated transformer

  • Representation Learning: A Review and New Perspectives

  • On the Opportunities and Risks of Foundation Models

Beyond text

• Week 12 (Nov 24). Language grounding: language+vision/robotics, pragmatics, RL agents

  • slides recording 1 recording 2

Conclusion

• Week 13 (Dec 1). Guest lecture by Nasrin Mostafazadeh: “How far have we come in giving our NLU systems common sense?”

  • slides recording

  • writing and presentation tips

• Week 14 (Dec 8). Project presentations