University of Waterloo >
Electronic Theses and Dissertations (UW) >
Please use this identifier to cite or link to this item:
|Title: ||CBKR+: A Conceptual Framework for Improving
Corpus Based Knowledge Representation|
|Authors: ||Ivkovic, Shabnam|
|Keywords: ||Computer Science|
Corpus-based Knowledge Representation
|Approved Date: ||2006 |
|Date Submitted: ||2006 |
|Abstract: ||In Corpus Based Knowledge Representation [CBKR], limited association capability, that is, no criteria in place to extract substantial associations in the corpus, and lack of support for hypothesis testing and prediction in context, restricted the application of the methodology by information specialists and data analysts. In this thesis, the researcher proposed a framework called CBKR+ to increase the expressiveness of CBKR by identifying and incorporating association criteria to allow the support of new forms of analyses related to hypothesis testing and prediction in context. |
As contributions of the CBKR+ framework, the researcher (1) defined a new domain categorization model called Basis for Categorization model, (2) incorporated the Basis for Categorization model to (a) facilitate a first level categorization of the schema components in the corpus, and (b) define the Set of Criteria for Association to cover all types of associations and association agents, (3) defined analysis mechanisms to identify and extract further associations in the corpus in the form of the Set of Criteria for Association, and (4) improved the expressiveness of the representation, and made it suitable for hypothesis testing and prediction in context using the above.
The application of the framework was demonstrated, first, by using it on examples from the CBKR methodology, and second, by applying it on 12 domain representations acquired from multiple sources from the physical-world domain of Criminology. The researcher arrived at the conclusion that the proposed CBKR+ framework provided an organized approach that was more expressive, and supported deeper analyses through more diagnostic and probability-based forms of queries.
|Department: ||School of Computer Science|
|Degree: ||Master of Mathematics|
|Appears in Collections:||Electronic Theses and Dissertations (UW)|
Faculty of Mathematics Theses and Dissertations
This item is protected by original copyright
All items in UWSpace are protected by copyright, with all rights reserved.