Trish Heiman
Summary of Monday's class (9/11)

I would like to have spent more time on Inverted file systems.  I am a
little behind in my reading; hopefully the book will clarify.

Pat Trees seem to provide a logical way of organizing ?addresses? with
as little memory as possible.  Hopefully next week's class will clarify
the connection to document storage and retrieval.

... I thought the debates for the Digital Library unit were
valuable.  However, the groups need to be separated; it was just too
noisy and distracting.  If we debate again, I would suggest having us
go off for 30-45 minutes and then come back and have general discussion
with you.

Also, I love the class notes on the WWW.  Because of these notes and
summaries, I don't feel like I have to take notes in class and
therefore I can listen more closely.  I would prefer more examples and
discussion and less reading of the bulleted items.  Maybe we could
solicit examples from students and promote more class interaction.

= = = = = ==  =  = === =  = =  = = = = =   =  ====  = = == =  = =  == = =

 class summary for the week of 9/11.

Carolyn O'Hare
Lauren Barton
Robert Ryan
Martin Falck
Nelson Kile
 
 
        We began class with a discussion of  the problems with Boolean Queries.
 Some of the problems with Boolean Queries are that they do not allow
ranking, weights on query terms or weights on document terms.  The different
Extended Boolean Models were examined.  These models include fuzzy set
theory, MMM Model, Paice Model and P-Norm Model.
 
Fuzzy Set Theory
 
        -       uses a range from 0 - 1 instead of a choice of 0 and 1
        -       redefines AND as MIN
        -       redefines OR as MAX
        -       evaluates NOT B as 1 - value(B)
 
MMM Model
 
        -       redefines AND and OR as linear combinations of MIN and MAX
        -       AND  = Cand * MIN + (1-Cand) * MAX
        -       OR    = Cor * MAX + (1-Cor) * MIN
 
Paice Model
 
        -       considers all terms in the query
        -       uses a normalized geometric series
        -       uses a single coefficient, r,  which is 1 for AND and .7
			for OR queries
        -       sorts document terms based on weights

P-Norm Model

        -       consider all terms in query
        -       paramerterize the strictness of each AND or
			OR operator with a p-value

        In comparison to Boolean, all the other models are more expensive than
Boolean but are also more effective.  Of the other Models MMM is
computationally cheaper and more effective than Paice and P-Norm.

        In the second part of the class, we began discussing string
searching and
PAT trees.  In string searching, the three types of text searching are
cluster trees, hashing and sorted indexes.

        With PAT trees, we learned how to read and traverse a PAT tree.
PAT  trees
were defined as a binary digital Patricia tree of all sistrings (semi-infinte
strings) in a text.

        The different algorithms that are performed on PAT trees are:

        -       Prefix searching
        -       Proximity searching
        -       Range searching
        -       Longest repetition searching
        -       Most frequent searching
        -       Regular expression searching