The Hong Kong University of Science and Technology
Department of Computer Science


PhD Thesis Defence


"On the Use of Web Data Quality Dimensions for Finding High Quality Web
Pages"

By

Mr. Chun Chung Pun


Abstract

While users can readily find information from the immense store of
knowledge in the Web with the help of a search engine, they often complain
about the quality of the results returned. The quality of a web page is
not necessarily only limited to its relevance, as judged by most search
engines, but can also depend on other features of a web page.

This thesis first examines what factors constitute a high quality web
page; hence a framework of web data quality dimensions is proposed. From
this framework, it is found that current search engines only consider a
very small number of web data quality dimensions. We then propose a
general methodology for evaluating web data quality metrics derived from
the web data quality dimensions. The methodology has been applied to two
web data quality dimensions (that is, appropriateness and cohesiveness) as
well as the combination of these two dimensions. Metrics were developed to
measure each dimension from a web page. They have been verified to measure
users' expectations.

The web data quality dimension, appropriateness, measures how well the
results returned from search engines satisfying the web genre needs of a
user. It is based on the linguistic and visual complexity of a web page.
The web data quality dimension, cohesiveness, is a measure of how closely
the concepts within a web page are related to each other. A distance
metric is defined to measure how close two concepts are in an ontology and
the cohesiveness of a web page is calculated as the total distances of all
the concepts within it. In addition, a technique to combine different
quality metrics and to incorporate user's preference for each web data
quality dimension is proposed. With these metrics, users can more easily
find high quality web pages (i.e., not just relevant web pages, but also
web pages matching other desired web data quality dimensions).

A user evaluation has been conducted to show the conformance of the
metrics to the corresponding web data quality dimensions (i.e.,
appropriateness, cohesiveness and combined dimensions). It also revealed
that the judgments on these dimensions from users are fairly consistent.


Date:			Thursday, 19 January 2006

Time:			10:00a.m.-12:00noon

Venue:			Room 4480
			Lifts 25-26

Chairman:		Prof. Songnian Chen (ECON)

Committee Members:	Prof. Frederick Lochovsky (Supervisor)
			Prof. Wilfred Ng
			Prof. Vincent Shen
			Prof. Richard So (IELM)
			Prof. Carolyn Watters (Comp. Sci., Dalhousie Univ.)


**** ALL are Welcome ****