Assignment 3: Parsing

Assigned: February 27th, 2006
Due: March 5 , 2006

Specifications:

Goal: This is an exercise on NLP parsing.

1) Obtain a CFG Parser.

Here are some pointers (there are many more resources out there):

                 Stanford Parser (Manning and Klein)
                 Collins Parser   Statistical Lexicalized Grammar.
                 Charniak's Parser Statistical Context Free Grammar
                 NLTK   Natural Language tools in python, it has the CFG model and Parser. You may want/have to specify your
                               own grammar.

You are free to use also non NLP parsers (deterministic or non deterministic, e.g. Earley's, Lex & Yacc), however this will amount probably to more work (the idea is that you do as much experimentation as you want).

In your submission, briefly describe what kind parser it is.

2) Parse half of your corpus from assignment 2 (e.g. aprox 250 words).

Submit the parsing output.

3) Write a script to recognize Explicit Subjects and Objects of a sentence (main or embedded sentence) from the parser output, all the other arguments are ignored. (If there is a relative clause, leave the relative clause out of the argument).

             Examples:    <subject John> visited <object England>.
                                     < subject Mary> bought <object a book> for John.
                                     <subject Mary > bought <object John > a book. (note we select here just one object).
                                      <subject I > found < object the book> that <subject Mary> bought
Submit the output of your script.

4) Score how well the recognition does, in terms of Precision/Recall (Exact Match for each of the Arguments).