December 11, 2002
Take Home Final Exam - CS 8761

Concept specificity is a hypothetical measure that provides a numeric  
value for each sense in a dictionary. This indicates how general or   
specific the concept associated with that sense is. Assume that more  
specific concepts have higher scores than more general concepts. For  
example, the concept associated with the sense "four wheeled motorized  
passenger  vehicle" (automobile) might have a score of 4.5, while the 
concept associated with sense "solid entity that occupies space and has  
mass"  (object) might have a score of 1.0.

Part A (500 words or less) 

Suppose that we want to estimate these values using two resources: a huge  
billion+ word corpus, and the dictionary that is being augmented. How  
could you use these sources of information to arrive at values of  
"concept specificity" for each sense enumerated in that dictionary?  

You may use the corpus and dictionary in any way, but DO NOT assume the 
availability of other knowledge sources or specialized tools/techniques  
beyond what we have already discussed or used in the class. Assume that 
you only have one dictionary available, and that it is either BigMac, 
LDOCE, or some idealized version of either one of these. 

Part B (500 words or less)

Once calculated, what could you do with these values? Give an example 
of a problem that could be better solved if you had values of concept  
specificity. If you can't think of a problem that could be solved, 
describe some interesting relationships among words we could determine   
based on this sort of information. 

-----------------

Please write no more than 1000 words in total! If you write more than 
than 500 words on Part A and/or Part B you will receive a 5% reduction 
in your grade for each offense. So if both parts are overly long you
will lose 10%. I will also make deductions for obvious spelling and
grammar errors.  Submit your response via the web drop by noon on Wed  
Dec 18. Please submit as a plain text file that is named <userid>.txt.

Your response will be judged more for its insight and creativity than its  
feasibility or perhaps even its correctness. 

You may not consult any other human being in formulating your response.  
If you utilize ideas from published sources, please cite them appropriately. 
Do not plagiarize. You should know what this means by now, and if you 
don't, refer to my Plagiarism Case Study, available on my home page. 
Please write clearly, and write well.