Homework 1 

Natural Language Processing
(CS 3730 / ISSP 3120)
Fall 2001

Assigned:  September 12, 2001
Due: September 26, 2001

1. (10 pts) Amelia Bedelia is a "literal-minded but charming
housekeeper who never fails to confound the Rogers family" (from the
jacket of Amelia Bedelia, by Peggy Parish).  For example, when told to
"Draw the drapes when the sun comes in," Amelia waits for the right
time of day, takes out her pad, and sketches the drapes.  What do you
think Amelia does when told to "Dress the chicken" after it is
delivered by the meat market?  Explain why this book has kept children
laughing, using the parlance of Chapter 1.

2. Consider recognizing a simple numerical language, consisting of
integers and real numbers, where there is an optional preceding plus
or minus, and a required digit before the decimal point.  Leading
zeros are not ok, but trailing zeros are. Thus 12, 3.110, 12.34, +12,
-12.34, 0.1 are in the language, but .1 and 01 are not.

a) (10 pts) Design a NFSA using epsilon transitions to recognize this language.
b) (10 pts) Design a DFSA that removes the epsilon transitions.
c) (10 pts) Write a regular expression using only the primitive operators to recognize this language.

- You may use some sort of shorthand for classes of symbols, e.g. in a 
furniture domain, furniture instead of desk, chair, table

- If you write your FSAs in a table representation, please also
draw the equivalent graph version (by hand is fine)

3. (60 pts) Implement the ND-Recognize algorithm from Jurafsky and
Martin (page 44) using depth first search. Test your algorithm using
sheep language NFSA#1 (Fig. 2.18 in Jurafsky and Martin), as well as
the NFSA developed for question 2 above.  For each test, show several
examples where your algorithm accepts an input string, and examples
where it does not.

- Your code should be table-driven (i.e., represent the FSA as a table).
- Use whatever programming language you like.
- Don't worry about out of alphabet symbols.
- Your output should include the input and outputs of the algorithm.
- Also, your output should trace the execution of the machine to illustrate the handling of backup.