Computer Science 248/448
Statistical Speech and Language Processing
Fall 2020
|
|
Instructor:
Dan Gildea (Office hours: Mon 2:00-3:00pm, Thu 2:00-3:00pm)
TA: Will Gantt (Office hours: Tu/Th 4:00-5:00pm)
Time: Tu/Th 11:05am-12:20pm
Homeworks
Recommended resources:
Syllabus
| On | we will cover | which means that after class you will understand | if before class you have read |
| 8/27 |
Perceptron for Tagging |
NLP = ML + DP
| Collins 2002 |
| 9/1 |
Hidden Markov Models |
viterbi, forward
| JM App A |
| 9/3 |
CRFs |
f - E[f]
| McCallum 2001 |
| 9/8 |
Neural CRFs |
convexity
| Peng 2008 |
| 9/10 |
Context Free Grammars |
highest prob string, infinite trees
| JM 13 |
| 9/15 |
latent CRF |
| |
| 9/17 |
loss-aware CRF |
| Kakade 2002 |
| 9/22 |
EM for HMM |
| JM App A |
| 9/24 |
Context Free Grammars |
viterbi, inside-outside
| |
| 9/29 |
Context Free Grammars |
viterbi, inside-outside
| |
| 10/1 |
Parsing |
minimum risk decoding, max-margin
| Finkel 08 |
| 10/6 |
Parsing |
crf parsing, neural parsing
| Durrett 2015 |
| 10/8 |
Weighted Deduction |
generalized outside computation
| Goodman 1999 |
| 10/13 |
Review |
| |
| 10/15 |
Midterm |
|
|
| 10/20 |
Midterm Solutions |
| |
| 10/22 |
Pytorch |
tensors
| tutorial |
| 10/27 |
Search |
dijkstra, knuth, a*
| |
| 10/29 |
Beam search
| buckets
| |
| 11/3 |
no class
|
| |
| 11/5 |
Machine Translation |
Model 1
| Brown 1993 |
| 11/10 |
Neural Machine Translation |
attention
| Sutskever 2014, Bahdanau 2014 |
| 11/12 |
Neural Machine Translation |
BPE, BLEU
| Vaswani 2017, Papineni 2002, Sennrich 2016 |
| 11/17 |
Semantics |
word2vec
| Mikolov 2013, Pennington 2014 |
| 11/19 | Encoder Pretraining |
bert and elmo
| Elmo, Bert |
| 11/24 |
Semantics |
lambda calculus
| JM 20, Zettlemoyer 2005, Liang 2011 |
| 12/1 |
Something fun |
| |
| 12/3 |
Something fun |
| |
| 12/8 |
Something fun |
| |
Grading
- Term project: 35%
- Homeworks: 40%
- Midterm: 25%
- No late homework
- Programming assignments must be in python, and must run on class account
gildea @ cs rochester edu
November 2, 2020
|