| Topic
| Readings
|
Language modeling and EM
1/23,
1/25
|
|
Log-linear models
1/30
2/1
|
|
Neural models
(2/6,
2/8),
Recurrent NNs
(2/13,
2/15)
|
- Ch. 2, Eisenstein (2018) book draft, "Nonlinear classification"
- Bengio et al. (2003), on a Markov multilayer perceptron language model - reading for Feb 6.
- Linzen et al. (2016), Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies - reading for Feb 13.
Optional:
|
|
Distributional lexical semantics
(2/20, 2/22)
|
(No reading assignment this week.)
|
|
Statistical testing and annotations
(2/22,
2/27, 2/29)
bootstrapping demo
|
|
|
Sequence Labeling:
POS, HMM
(3/6,
3/8,
HMM handout)
|
- Eisenstein (2018) draft: Chapter 6, "Sequence labeling"
- Eisenstein (2018) draft: Chapter 7, "Applications of sequence labeling"
- Optional:
Johnson (2007),
Why doesn’t EM find good HMM POS-taggers?
|
|
Spring break
|
|
Forward-Backward
(Week of 3/20)
|
|
|
Conditional random fields
(3/27,
CRF example,
3/29,
Viterbi worksheet)
|
- Eisenstein chapters above.
- Also, paper reading for Thursday. It was just selected
for a Test of Time award:
-
Collins (2002).
Discriminative Training Methods for Hidden Markov Models:
Theory and Experiments with Perceptron Algorithms.
Proceedings of EMNLP.
|
|
Constituency Parsing
(4/10, 4/12)
|
- Eisenstein (2018) draft: 8.2 on Context-free languages
- Eisenstein (2018) draft: Chapter 9, "Context-free Parsing"
|
|
Note no class on 4/17 (Monday schedule day).
|
|
RNNGs
(4/19)
|
|
|
Ethics and NLP
(4/24)
|
|
|
Guest lecture
(4/26)
|
- Keith et al. (2017),
Identifying civilians killed by police with distantly supervised entity-event extraction. EMNLP.
|
|
In-class presentations:
5/1
|
|