Introduction to Natural Language Processing

CSCI 490-B1/680-A6

Fall 2009

Last updated Thu. 11/11/2009

Back to Reva Freedman's home page

Announcements

• Additions and corrections to this page are always welcome! Please send improvements and corrections to the lecture notes.
• Regular office hours are TTh 3:30 PM-5:15 PM, or by appointment (usually earlier or later on TTh, sometimes on Wed., occasionally on Fri.
• Schedule for last part of semester

Format

Some documents on this page are in PDF format. One way to read them is to download a free copy of Adobe Acrobat Reader. If you already have a program to read PDF but the documents don't automatically pop up when you click on them, you need to configure your browser.

If you would like to print multiple slides per page, check your printer driver. This option is often available under the "layout" tab or something similar.

Syllabus

Here is the syllabus for this course (PDF).

Resources

NLTK

• Book
• Google code browser

Python official web site

• Python package repository Google is sometimes more useful.
• Python documentation for version 2.5.2
• Python tutorial for version 2.5.2

More Python

• A Python programmer's summary of NLTK
• TKInter

Dialogue Systems

• A very simple Eliza
• Alicebot blog: chat with Alice, chat with God, chat with Santa
• Alice talks to Jabberwocky
• Eliza talks to Parry, the vice president of BBN, etc.

NLP

• Jurafsky and Martin textbook web site
• ACL Anthology: Digital Archive of Research Papers in Computational Linguistics
• LingPipe: Java Libraries for NLP
• Language Experiments

Linguistics

• Ethnologue: Languages of the World

Artificial Intelligence

• AI in the News (from AAAI)

Assignments

Assignments must be submitted with the submit script for this class. No late homework. Keep your receipt: it is your only proof of submission. If you get a vacation message (whether or not you get a receipt), you have missed the deadline.

• Submission instructions

• HW1: Making sure you can access Python. Due midnight, Tue., Sept. 29.
• HW2: Basic data structures and control structures Due midnight, Thu., Oct. 1. A test file for HW2
• HW2a: A cleaned up (and slightly longer) version of HW2 Due midnight, Fri., Oct. 2. You can do this one instead of HW2 if you prefer, i.e., if you want the extra day, you need to do this version.
• HW3: Finite state machines, GUIs, and parts of speech. Due midnight, Thu., Oct. 22.
• HW7: Required project for graduate students. Due midnight, Mon., Nov. 30.

Notes

Notes from 2008

• Linguistic background
  • International Phonetic Alphabet (IPA): Language Files 3.1 and 3.5
  • English consonants: Language Files 3.2
  • English vowels: Language Files 3.3
  • Phones vs. phonemes: Language Files 4.1


• What makes NLP hard
  • Language variation: Language Files 10.10
  • Surface form vs. meaning (the duck sentences)
  • Types of ambiguities
  • Even knowing what's a word is hard
  • More types of ambiguity
  • Humans have problems too: Garden path sentences
  
  
  • English morphology
  • Evidence for human two-level processing
  • Stemming without a lexicon (Porter)


• Finite-state automata
  • Old FSA HW


• Spelling correction
  • Bayesian spelling correction: spell09a.pdf
  • Old spelling HW
  • Kernighan's COLING '90 paper
  • AP '88 unigram and bigram counts
  • Bayes' theorem for ASR


• TBD
  • Pragmatics: Intention vs. form

Preliminary

• CFGs
• Greenberg's universals
• Language change over time
• Meaning