Class Summary, Sep 26 -- Srinivas R. Gaddam

Dr.Fox mentioned that he included example sessions for searching
pattern in text using the Naive and Boyer-Moore algorithms. Also,
similar such examples will be included for other algorithms. 

Dr.Fox gave an overview of Marian, a search system for library catalog data.
It is being developed as an alternative to VTLS, the system currently used
in Newman library. One of the many features Marian provides is that one can
list the results retrieved in the decreasing order of relevance. 
 
Then the lecture continued with discussion on Probabilistic Model. In this
model, document is represented by a binary vector x and not as a single entity
D. This binary vector consists of a set of 0's and 1's. A 0 or 1  in the ith
position indicates the absence or presence of the ith index term. 
Using Bayes Theorem, one can calculate the relevance or non-relevance of a
document given x.  This model is advantageous over the term dependence models
because in the latter case, a function is applied and the query is not taken 
into account later on. Also, features like term frequency effect and weights
can be added to the probabilistic model.

Then, the topic of Inference Networks was introduced. An example that uses
the concept of Inference Networks is the INQUERY Retrieval System. This system
can be used to search large databases like the library of congress collection 
in a very short time. Various features of INQUERY include fast retrieval time,
relevance feedback, good recall and precision.

Topics on Evaluation, Data Structures, Optimization were also briefly
discussed.