assn1

Use NLTK to implement the grammar you defined in exercise 1. If you find it useful, you may use the example code in extraction_grammar.py in the class code directory as basis for writing and testing your grammar. You should make sure that the good sentences are all parsed and that the bad sentences receive no parse. (The example code will make this easy.) You should provide output similar to that found in the example code.

You may find that you need to revise your grammar as the implementation exposes flaws in it (eg, it can parse sentences that should be disallowed). Some of these may be implementation bugs while others may be thinking bugs that were actually problems with the grammar you defined in exercise 1. Comment on how this process went for you, especially if the implementation revealed the latter type of error. (Note: it is fine to change your answer to exercise 1 if the implementation did expose errors!)

2. Finding subcategorization frames

Choose a text collection (one provided by NLTK, the POS tagged 9/11 Report, or any other you may want to use, either from the data directory or from any source you can find on the web or elsewhere. If your data is not already POS tagged, you should run a POS tagger over it first.
Choose a verb that interests you, and find all the sentences that are tagged with that verb (probably best to use all of its conjugated forms). Be sure you have a good size number of sentences containing the verb in its various forms (at least 30).
Using the NLTK shallow parsing facility (chunk, unchunk, and chink rules along with RegexpChunkParser), produce shallow parses of the selected sentences. You may want to start with the exmaple chunkers presented in class, but you should improve on them greatly. I suggest using multiple rules for each type of chunk (NP, VP, PP, others if you like). When you turn in the assignment you should show some before and after parses on the same sentences to show how much your rule rewrites have improved the chunker.
Analyze the argument structure of the verb that you've chosen. What kinds of subject and objects does it tend to take, both syntactically and semantically (e.g. in terms of animacy, etc.)?
(Optional) Try to find at least two verbs that take objects or subjects that are similar in form, either syntactically, semantically, or both. If you can't find any, try to describe why not. Be sure to describe how you tried to find the similar verbs.
(Optional) Use the rules you developed to parse sentences containing the verb you chose that are taken from a text of a different genre and/or time period. For example, if you chose Jane Austen you might try on the NY Times or 9/11 Report. Do your rules still work? If not, why not? Can you augment or modify the rules to work on the new text?

bigrams	computes bigrams for a corpus
buildStoplist	buildStoplist : builds a list of stop words (words not to be included in an analysis)
compareModals	comparison of modals in parts of the Brown corpus (from class slides)
concordance.py	generates a concordance of a given word as well as collocates
conditional_prob	computes a conditional probability for a POS tag for each word
countTags	computes the tag most often assigned to words in a tagged corpus
find_names	finds proper names (sort of) (from class slides)
freqdist	computes a frequency distribution for words in genesis and prints the token with greatest number of occurrences (from class slides)
init	reads the plain text 911 report file
storyGen	try to generate a story (from class slides)
wordcounts	counts words in the text
wordcounts-text	does the same as wordcounts but prints each word and frequency
extraction_grammar.py	Defines a grammar and parses supplied sentences

Assignment 3 : Grammars and Parsing

Exercises

1. Creating and implementing a grammar fragment

Resources

Code

Links