Please use this identifier to cite or link to this item:
Title: Relation extraction for ontology extension using integrated evidences
Authors: Chen, Yirong
Degree: Ph.D.
Issue Date: 2011
Abstract: Ontology is a valuable resource for many domain specific applications where domain knowledge is needed. With the rapid development in science and technology, new terminology and associated concepts must also be updated in the ontology to suit for real time applications. Current methods of manual construction of ontology is too time consuming and difficult to update. Thus automatic extension of ontology is especially needed. In this study, investigation to terminology extraction is first carried out. In addition to unit-hood measurement, this work further studies how to take domain specific knowledge to further measure term-hood to improved terminology extraction algorithms. After a thorough review of existing ontology resources, this study further investigates how to map the extracted terminology to a domain specific ontology. Given a core ontology, the key issue is how to find the relationships of the new terms to the concepts of the ontology. The investigation focuses on the extraction of kind-of relations. The work is divided into three steps: (1) To design effective algorithms to extract terms from domain corpus with good accuracy; (2) To investigate effective techniques to extract relations between concepts especially kind-of relations; (3) To link obtained ontology to upper ontology. The contributions of this work are three folds: (1) An effective term extraction algorithm is proposed based on the measures of both linguistic unit and domain specificity; (2) Algorithm of relation extraction is designed to construct domain ontology using multiple evidences; and (3) the construction and mapping of a core ontology to the upper ontology to ensure interoperability with other domain ontologies.
Subjects: Ontologies (Information retrieval)
Data mining
Knowledge acquisition (Expert systems)
Hong Kong Polytechnic University -- Dissertations
Pages: x, 235 p. : ill. ; 30 cm.
Appears in Collections:Thesis

Show full item record

Page views

Last Week
Last month
Citations as of Oct 1, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.