Advanced Campus Services
Information Systems & Technology
Georgia State University
P. O. Box 3968
Atlanta, Georgia 30302-3968
Phone +1 404 463 9685
Email: avandenberg@gsu.edu
Directory Services Project
October 25, 2002
1:30 pm - 3:00pm
Classroom South 514
Next meeting: Friday November 8, 2002- 1:30-3:00pm
Agenda Item 2: LSA/LSI demo - Jorge Civera
Hi,
My idea for my LSA presentation tomorrow is to show a demo of LSA using MATLAB. I hope doing a small real example any possible doubt disappears.
The example is in the introductory papers I recommended about LSA. I would like to show how to:
Obtain a document by document matrix (object by object matrix)
Obtain a term by term matrix (attribute by attribute matrix)
Queries in LSA. This is a new issue I just pointed out in the previous meeting. The scenario I am thinking about is that a user wants to create a new object, but the user doesn't know if this object already exists in the Directory Service. The user will provide the attributes he wants his new object contains. Then, the system will give as a result a ranking list of the most similar objects (already created) to that new one in the Directory Service. Goal: reuse of objects and homogeneity.
Have created the database schema (tables) in the Oracle database for storing the LDAP directory information.
I have used the LDIF files I had received from you/Victor to store the directory information into the database.
I am currently looking at what kind of indexes (if any) we should have on these database tables.
For storing the intermediate results of the SOM algorithm, I would like to request some input from you all: are we interested in storing the matrix in the database as is, or do we want to annotate it with some useful information before storing it into the database. Also, would the LSA we are looking at now cause any impact on this? Depending on these factors, I would be able to proceed further.
