Crisp Representation of Vague Ontologies

From Master Projects
Revision as of 09:39, 16 December 2008 by Mcaklein (talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search


has title::Crisp Representation of Vague Ontologies
status: finished
Master: project within::Knowledge Technology and Intelligent Internet Applications
Student name: student name::Ren Yuan
number: student number::1668773
Dates
Start start date:=2008/05/20
End end date:=2008/11/26
Supervision
Supervisor: Dr. Zhisheng Huang, Dr. Jeff Z. Pan
Second reader: has second reader::Stefan Schlobach
Poster: has poster::Media:Media:Posternaam.pdf

Signature supervisor



..................................

Abstract

The fast development of today's world wide web arouses the requirement of web knowledge representation and automatic information processing, which leads to the vision of the semantic web. However, ontologies, the knowledge infrastructure of the semantic web, are based on classical description logics. Therefore, they lack the ability of dealing with uncertain information, which are very common in the web environment.

In this project, we are particularly interested in the representation of vagueness, a specific type of uncertainty featured by its multiple truth values and frequent appearance in nature language. To solve this problem, we investigated the nature and source of various uncertainty phenomena to propose the requirements of vagueness representation in the context of the semantic web. We reviewed several mainstream uncertainty representations to compare the similarities and differences of their modeling principles. We conducted a comprehensive survey of major uncertainty extensions of description logics to argue that, a reduction from fuzzy description logics to crisp description logics provides a solution that can (1) reflect the features of vagueness phenomena; (2) achieve a balance between expressivity and scalability and (3) guarantee compatibility with existing semantic web data and tools. We further implemented the reduction approach to validate the scalability of the reduced ontology with experiments.

The results showed that currently the crisp reasoning upon the reduced ontology is significantly more scalable than the immediate fuzzy reasoning upon the fuzzy ontology, both in terms of terminological classification and in terms of query answering. The similarity of various uncertainty representation also suggested that such a reduction approach can possibly be generalized to facilitate the reasoning of other uncertainty description logics.