Login

Publications  •  Project Statistics

Glossary  •  Schools  •  Disciplines
People Search: 
   
Title/Abstract Search: 

Dissertation Information for Yuri Quintana

NAME:
- Yuri Quintana

DEGREE:
- Ph.D.

DISCIPLINE:
- Computer Science

SCHOOL:
- University of Waterloo, Ontario (Canada) (1995)

ADVISORS:
- None

COMMITTEE MEMBERS:
- None

MPACT Status: Incomplete - Not_Inspected

Title: Knowledge base organization using knowledge intensive conceptual clustering

Abstract: The ability to learn is one of the key components of human intelligence. This thesis examines factors that influence how humans create clusters of information, and introduces a new conceptual clustering algorithm that can be used to organize knowledge bases. The design of the conceptual clustering algorithm is based on experiments that required humans to cluster and classify a variety of stimuli. The stimuli were selected from published results of existing machine learning algorithms. Diverse clustering instructions were given to humans in order to analyze the effect of the instructions on the cluster descriptions. The clusters were also analyzed to see the language that was typically used and which features were most prominent. Based on these experiments, recommendations are made for the design of conceptual clustering algorithms that can create clusters similar to humans. A new conceptual clustering algorithm is presented that incorporates these recommendations. The algorithm uses generalization, background knowledge, and semantic similarity heuristics to create clusters and conceptual descriptions of their contents. New methods for comparing clusters that are created by both humans and machines are also described. Using these methods, the clusters created by the new algorithm were found to be similar to the clusters created by a sample group of test subjects. The new conceptual clustering algorithm was also used to organize a picture knowledge based that describes the semantic content of pictures. Picture retrieval experiments were conducted with sample queries collected from human test subjects. It is shown how a semantic-based search of an organized picture knowledge base is faster and more accurate than a sequential search of the knowledge base, and significantly more accurate than a keyword search of text captions.

MPACT Scores for Yuri Quintana

A = 0
C = 2
A+C = 2
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2012-07-30 16:34:04

Advisors and Advisees Graph