Class Plans for Ling 645/CMSC 723, Fall 1997
Class Plans for Ling 645/CMSC 723, Fall 1997
Introduction to Computational Linguistics
- September 3, 1997
- September 10, 1997
- September 17, 1997
- September 24, 1997
- October 1, 1997
- October 8, 1997
- October 15, 1997
- October 22, 1997
- Oct 29: no class, Linguistics Student Conference
- Nov 5, 1997
- Nov 12, 1997
- Nov 19, 1997
- Nov 26, 1997
- Dec 3, 1997
- Dec 10, 1997
Introduction and administrative stuff
What is computational linguistics?
Computational linguistics as science, engineering
Levels of natural language analysis
Assignment 0:
Log in to your course account on one of the aITs machines
Change your password!
Read your e-mail
Save an e-mail message to a file
Send "hello!" e-mail to pr01@umd5.umd.edu
Look through course page, http://umiacs.umd.edu/~resnik/ling645.html
Look at CL Colloquium page, http://umiacs.umd.edu/~resnik/cl_colloquium/
For those with little computing experience:
- Take the Peer Training course on Unix, and/or
- Run the 'learn' command from your course account and start
working your way through "files" and "editor", and/or
- Run the 'emacs' text editor, and go through the emacs tutorial
(invoked via "ctl-h t")
Reading for next week:
Allen, Ch 1-2, and get started on Ch 3
Assignment 1a:
- Invoke the 'emacs' text editor
- Run the "Eliza" program by typing "Meta-x doctor"
(An alternative: http://www-ai.ijs.si/eliza/eliza.html)
- Play with the program for a while: can you get it to
have a reasonably natural sounding dialogue for a sentence
or two? If so, what makes it natural? If not, what are
the obstacles? [Bring a written answer to class,
showing a dialogue or two you tried.]
Assignment 1b:
- Allen, Chapter 1: exercise 2.
- Allen, Chapter 2: exercises 2, 4.
[Bring written answers to class]
Solution set
Components in an NLP system
A real-world example
Finite-state automata: basics (not in Allen's book)
Reading for next week: continue with Allen, Chapter 3
Assignment 2:
Part 2a: Basics of LISP
Part 2b: Working with a DFA
Part 2c: An NDFA for the English Auxiliary System
Clarifications on assignment 2
Some additional Formal language theory notes
.
Partial solutions for Assignment 2
Nondeterminism via backtracking: NDFAs
Parsing as search using context-free grammars
Top-down, bottom-up, and left-corner parsing
Chart parsing: the CKY algorithm
Dotted rules and active edges
Reading for next week: Allen, Chapter 4
Assignment 3:
Part 3a: More basics of LISP
Part 3b: Modifying the NDFA simulator
Part 3c: CFG formalisms and parsing
Solution set for Assignment 3
Chomsky normal form; weak and strong equivalence
Left-corner bottom-up parsing
Dotted rules and active edges
Chart parsing: the Earley algorithm
Shieber's proof that natural language is not context free
Assignment 4: More on CFG parsing
Solution set for assignment 4
Beyond CFGs
Augmented transition networks (ATNs) as extending RTNs.
Augmenting grammars with features
Assignment: review for mid-term exam on Oct 8
start playing with Allen's parser (optional)
First midterm (90 minutes)
Beyond CFGs: tree-adjoining grammar
Reading for next week: Allen, Appendex A.1; start Chapter 8
Assignment 5: Using a CFG with features
Beginning semantics
Review of propositional logic and first-order logic
Discussion of word senses
WordNet
Assignment 6:
Exploring word senses using WordNet
Reading: Continue with Allen, Chapter 8
Lambda expressions and compositional semantic interpretation
Rule-to-rule correspondence of syntax and semantics
Event variables
Thematic roles
(Selection restrictions)
Assignment 7: Logical form and semantic interpretation
Solution set for assignment 7
Class Oct 29 cancelled for student conference.
Going over Assignment 7: semantics and logical form
Discussion of LF and QLF
- Predication/argument notation vs. Davidsonian event variables
- Syntactic ambiguity vs. semantic ambiguity: one QLF per parse
- Subcategorized vs. non-subcategorized prepositional phrases
Review of lambda-reduction
Verb lexical semantics
- The syntax/semantics relationship: diathesis alternations
- Illustration using the dative alternation
- Extended example from Levin book: hit, touch, cut, break
- Sketch of decompositional verb representation: LCS
- LCS as a language-independent representation
Midterm exam
Reading for next week: Allen Ch 10 Sec 1 through 3.
Reading for this week: Allen Ch 10 Sec 1 through 3.
Use of semantic and world knowledge in processing
- Selection restrictions
- Constraint satisfaction
- Use of selection restrictions in parsing
- Semantic networks
Assignment 8: selection restrictions and constraint satisfaction
Reading for this week from Allen:
Section 7.1-7.2
Section 7.3 excluding Viterbi algorithm
Section 7.5
- Intro to corpus-based methods
- N-gram models
- Probabilistic CFG
- Noisy channel model
Assignment 9: probabilistic models
Reading:
Church and Hanks (1990)
Hindle and Rooth (1993)
First half of class (bracketed indicates not covered)
- Noisy channel/HMM for part of speech tagging
- Evaluation issues: [training vs. test data], lower and upper bounds
- [Probabilistic disambiguation of PP attachment]
- [Mutual information and lexical association]
Second half of class
- Guest lecture on machine translation by Bonnie Dorr
Assignment 10: statistical methods
Reading: Lewis and Sparck-Jones (1996), "Natural Language Processing for
Information Retrieval"
First half of class
- Basics of information retrieval (IR)
- Evaluation issues: recall and precision
- Some attempts to improve IR using NLP
Second half of class
- Semester wrapup
Final exam is December 16
Return to the course home page.