Tony Mullen

Special Lecture in Computer Science A:

Natural Language Processing in Python

This course builds on much of the material covered in last term’s Introduction to NLP, but with an added emphasis on practical implementation. The course will also serve as an introduction to programming in Python. Topics in NLP covered will include tokenization, part-of-speech tagging, syntax and parsing, and statistical and probabilistic methodologies. Students will make use of the Natural Language Toolkit (NLTK) in Python and will be expected to do a good deal of programming.

Required texts include Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics and Speech Recognition by Daniel Jurafsky and James Martin and the NLTK tutorials.

Recommended texts include Dive Into Python by Mark Pilgrim and 初めてのPython―Object‐oriented programming by Mark Lutz and David Ascher.

A strongly recommended resource is the Python Tutorial by Guido van Rossum.

Here is a regular expression tutorial.

And here is some information in Japanese.

Week 1

●Read NLTK Tutorial: Preface and NLTK Tutorial: Introduction to Natural Language Processing

Run the three graphical demonstrations here:

http://nltk.sourceforge.net/getting_started.html

●In the recursive descent and the shift-reduce parser, you can edit the grammar and the text to be parsed

under “edit”. Edit the grammar to include “friend” as a noun, “bit” as a verb, and “my” as a determiner.

Parse the sentence “a dog bit my friend in the park” in each parser and take screenshots of the correct parse tree.

●Read Chapter 2 of “Dive Into Python” (or the first 2 chapters of the introductory Python textbook of your choice)

If you plan to work on your own computer during this course, please try to install NLTK.

Check your work here.

Week 2

●Understand the examples in the NLTK tutorial “Tokenizing Text and Classifying Words”.

●Read Chapter 3 of “Diving into Python” or the chapter on native data structures (lists, tuples, dictionaries) in your preferred Python textbook.

●Create a “tuple” of five countries names, and a “list” of their capital cities. Using two variables to represent these structures, create a “dictionary” which associates the countries in the tuple with the cities in the list, using a single line of code. Cut and paste the sequence of commands from IDLE into an email.

-Hint: you will probably want to use the dict() and the zip() functions, which you can find out about in the Data Structures chapter of the Python Tutorial.

●Do Exercises 1 and 2 in "Tokenizing Text and Classifying Words" of the tutorial. For the text, please use this text.

Check your work here.

Week 3

●Computer room tutorial session.

Work on Week 2 exercises.

Week 4

●Review Objects and Classes/Frequency Distributions.

●Do Exercises 3, 4 and 5 in "Tokenizing Text and Classifying Words" of the tutorial.

Solution to exercise 3.

Solution to exercise 4

Solution to exercise 5

Week 5

●Understand all examples in NLTK Tutorial:Tagging

●In NLTK Tutorial:Tagging do exercises 1 and 2

(Partial) solution to exercise 1

Week 6

●Computer room tutorial session.

Work on Week 4 and 5 exercises.

Week 7

●Review Chapter 9 in Jurafsky & Martin

●Read NLTK Tutorial: Chunking and understand the examples

●Do exercises 1, 2, and 3 in NLTK Tutorial: Chunking

Week 8

●Computer room tutorial session.

Work on exercises.

Week 9

●Do exercises 9.1, 9.3 and 9.5 in Jurafsky and Martin

●Read chapter 10 through 10.3

●In the NLTK parsing demo, create a simple grammar for japanese which is capable of accepting:

– inu ga neko o mita

– watashi ga hon o yonda

– watashi ha inu o mita

– neko ha inu ga mita

But not:

– *inu o neko o mita

– *watashi ga hon ga yonda

●Discuss the problems. What sentences are allowed which shouldn't be? Why?

●Consider how a grammar might be made to account for these phenomena using a "postpositional phrase" constituent consisting of an NP followed by a postposition. What difficulties come up?