CIS634 Retrieval Experiment 2

by Brook Wu

Credits:

• The test collection used was LISA.

• The experimental IR system was BOW toolkit, by Andrew McCallum.

• The multivariate analysis tool was the trial version of Nenet, a software program implementing the neural network algorithm -- Self-Organizing Maps (SOM).  SOM was developed by Dr. Teuvo Kohonen.

Query and Relevant Documents:

• The query for my experiment is Query 5:

"5
MY DISSERTATION IS AN OVERVIEW OF NEWS DATABASES, LOOKING AT THE AVAILABILITY, CONTENT, SCOPE AND USEFULNESS OF WHAT IS CURRENTLY AVAILABLE. IT WILL INCLUDE AN EXAMINATION OF USER GROUPS, BOTH AS PERCEIVED AND IN REALITY. MORE SPECIFICALLY, I SHALL CARRY OUT A BRIEF EXAMINATION OF THE HISTORY OF THE DEVELOPMENT OF NEWS DATABASES, STARTING WITH THE NEW YORK TIMES INFORMATION BANK AND ENDING WITH THE MOST RECENT DEVELOPMENTS AT REUTERS. NEWS, NEWSPAPERS, NEWS DATABASES, REUTERS, NEXIS, TEXTLINE, NEW YORK TIMES INFORMATION BANK, FINTEL, DATASOLVE, WORLD REPORTER. #"

• Relevant documents in the LISA text DB are:

Query 5
1 Relevant Refs:
3401

Part I: Data Collection (5%)

A Small Text Collection based on top 25 retrieved documents from model1 and model2:

• group1 (25 retrieved documents from model1)
• group2 (22 retrieved documents from model2)

Automatic Indexing:

• The vocabulary file for this test collection (with default settings: no stemming, stopped words removed)
• The original document-term matrix (with top 5 terms based on info-gain) 

SOM inputs and outputs:

• The document-term matrix prepared for SOM input
• The SOM matrix.cod output file
• The map:

Part II: What did you observe from the results of this experiment?(5%)

YOUR RESPONSE HERE!