Paul G. Allen School of Computer Science & Engineering

CSEP 517 - Natural Language Processing - Autumn 2018
Tu 6:30-9:20 in JHN 175

Teaching Staff

Personnel Contact
Instructor: Luke Zettlemoyer lsz at cs dot washington dot edu
TA: Srini Iyer (sviyer at cs dot washington dot edu)
TA: Mandar Joshi (mandar90 at cs dot washington dot edu)
TA: Julian Michael (julianjm at cs dot washington dot edu)
TA: Rowan Zellers (rowanz at cs dot washington dot edu)

Approximate Schedule [Subject To Change]


Week Dates Topics & Lecture Slides Notes (Required) Textbook & Recommended Reading
1 Oct 2 I. Introduction [slides]
II. Words: Language Models (LMs) [slides]
LM JM 4.1-4; MS 6
2 Oct 9 II. Words: Unknown Words (Smoothing)
III. Sequences: Hidden Markov Models (HMMs) [slides]
HMM JM 4.5-7; MS 6; JM 5.1-5.3; 6.1-6.5; MS 9, 10.1-10.3
3Oct 16
III. Sequences: Hidden Markov Models (HMMs) & EM Forward-backward, EM JM 6.6-6.8; JM 13-14; MS 11-12
4 Oct 23 IV. Trees: Probabilistic Context Free Grammars (PCFGs) [slides] PCFG
5 Oct 30 IV. Trees: PCFG Grammar Refinement
V. Learning (Feature-Rich Models): Log-Linear Models [slides]
Lexicalized PCFG,
LogLinear
Inside-outside
6 Nov 6 V. Learning (Feature Rich Sequence Models): Maximum Entropy Markov Models (MEMMs), Conditional Random Fields (CRFs), etc. [slides] MEMMs, CRFs
7 Nov 13 V. Learning (Feature Rich Sequence Models): continued
VI. Deep Learning: Intro Neural Networks [slides]
Feed Forward NNs
8 Nov 20 VI. Deep Learning: Recurrent Neural Networks (RNNs) [slides]
VII. Semantics: Distributed Semantics, Embeddings [slides]
JMv3 Vector Semantics,
9 Nov 27 VII. Semantics: Distributed Semantics, Embeddings (continued)
VII. Semantics: Frame Semantics [slides]
Frame Semantics,
JM 19.4; JM 20.7
10 Dec 4 VIII. Discourse: Coreference Resolution [slides]
IX. Translation: Neural MT [slides]
IBM Models 1 and 2, Phrase MT JM 25; MS 13

Textbooks

Assignments, Discussion Board

Available at Canvas

Grading

The grade will consist of four assignments worth 25% each (containing both written & programming problems). We will also award up to 5% extra credit for participation (in class or on discussion board).

Course Administration and Policies