Learning Ontology Relations by Combining Corpus-Based Techniques and Reasoning on Data from Semantic Web Sources

• Record Nr.: UNINA9910297042203321
• Autore: Wohlgenannt Gerhard
• Pubbl/distr/stampa: Frankfurt am Main : , : Peter Lang GmbH, Internationaler Verlag der Wissenschaften, , 2011; ©2011
• ISBN: 3-631-75384-5
• Edizione: [First edition.]
• Descrizione fisica: 1 online resource (222 pages)
• Collana: Forschungsergebnisse der Wirtschaftsuniversitaet Wien.
• Soggetti: Social ethics - Information technology; Business enterprises; Computer software
• Lingua di pubblicazione: Inglese
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Nota di contenuto: Cover -- 1 Introduction -- 2 The Semantic Web -- 2.1 Overview -- 2.1.1 Background and Vision -- 2.1.2 Features -- 2.1.3 Misconceptions and Criticism -- 2.2 Applications -- 3 Ontologies -- 3.1 Fundamentals -- 3.1.1 Purpose -- 3.1.2 Structure and Entities -- 3.1.3 Ontology Research Fields -- 3.2 Representation -- 3.2.1 Resource Description Framework -- 3.2.2 RDF Schema -- 3.2.3 Web Ontology Language -- 3.3 Querying and Reasoning -- 3.3.1 SPARQL and RDQL -- 3.3.2 Reasoning with Jena -- 3.3.3 Redland -- 3.4 Public Datasets and Ontologies -- 3.4.1 DBpedia -- 3.4.2 Freebase -- 3.4.3 OpenCyc -- 4 Methodology -- 4.1 Ontology Learning -- 4.2 Methods for Learning Semantic Associations -- 4.2.1 Natural Language Processing Techniques -- 4.2.2 Lexico-syntactic Patterns -- 4.2.3 Relevant Statistical and Information Retrieval Measures and Methods -- 4.2.4 Machine Learning Paradigms -- 4.3 Literature Review -- 4.3.1 Domain Text and Semantic Associations -- 4.3.2 The Web and Semantic Associations -- 4.3.3 Domain Text and Linguistic Patterns -- 4.3.4 The Web and Linguistic Patterns -- 4.3.5 Semantic Web Data and Reasoning -- 4.3.6 Selected Work from SemEval2007 -- 4.3.7 Learning of Qualia Structures -- 4.4 webLyzard Ontology Learning System -- 4.4.1 System Overview -- 4.4.2 Major Components of the Framework -- 4.4.3 ; Identification of the Most Relevant Concepts -- 4.4.4 Concept Positioning and Taxonomy Discovery -- 4.5 A Novel Method to Detect Relations -- 4.5.1 Relation Labeling Based on Vector Space Similarity -- 4.5.2 Ontological Restrictions and Integration of External Knowledge -- 4.5.3 The Knowledge Base -- 4.5.4 A Hybrid Method for Relation Labeling -- 4.5.5 Integration of User Feedback -- 4.6 Implementation of the Method -- 4.6.1 Training -- 4.6.2 Compute Vector Space Similarities -- 4.6.3 Ontological Restrictions and Concept Grounding -- 4.6.4 Scarlet.; 4.6.5 Evaluation -- 5 Results and Evaluation -- 5.1 Domain Relations and Domain Corpus -- 5.2 Evaluation of the Vector Space Model -- 5.2.1 Evaluation Baselines -- 5.2.2 Configuration Parameters -- 5.2.3 Average Ranking Precision -- 5.2.4 First Guess Correct -- 5.2.5 Second Guess Correct -- 5.3 Concept Grounding -- 5.4 Scarlet -- 5.5 Evaluation of Integrated Data Sources -- 5.5.1 Average Ranking Precision -- 5.5.2 First Guess Correct -- 5.5.3 Second Guess Correct -- 5.5.4 Individual Predicates -- 5.5.5 Summary and Interpretation -- 6 Conclusions and Outlook -- Bibliography.
• Sommario/riassunto: The manual construction of formal domain conceptualizations (ontologies) is labor-intensive. Ontology learning, by contrast, provides (semi-)automatic ontology generation from input data such as domain text. This thesis proposes a novel approach for learning labels of non-taxonomic ontology relations. It combines corpus-based techniques with reasoning on Semantic Web data. Corpus-based methods apply vector space similarity of verbs co-occurring with labeled and unlabeled relations to calculate relation label suggestions from a set of candidates. A meta ontology in combination with Semantic Web sources such as DBpedia and OpenCyc allows reasoning to improve the suggested labels. An extensive formal evaluation demonstrates the superior accuracy of the presented hybrid approach.