【12/14(木)17時より】CiNiiの常時SSL化(HTTPS接続)について

Theory and implementation of object oriented semantic Web language オブジェクト指向セマンティックWeb言語の理論と実装

この論文にアクセスする

この論文をさがす

著者

    • 小出, 誠二 コイデ, セイジ

書誌事項

タイトル

Theory and implementation of object oriented semantic Web language

タイトル別名

オブジェクト指向セマンティックWeb言語の理論と実装

著者名

小出, 誠二

著者別名

コイデ, セイジ

学位授与大学

総合研究大学院大学

取得学位

博士 (情報学)

学位授与番号

甲第1425号

学位授与年月日

2011-03-24

注記・抄録

博士論文

&nbsp;&nbsp;Resource Description Framework (RDF for short) is an assertional language intended to be used to express propositions using precise formal vocabularies, and its syntax is applicable to the WorldWide Webs. RDF Schema (RDFS) is a semantic extension of RDF and it provides a minimum typesystem to describe web ontologies. OWL Web Ontology Language is a language for defining andinstantiating Web ontologies. These three languages for Semantic Webs are intended to beintegrated in Semantic Web Layered Architecture, namely OWL was designed to be realized on topof RDF and RDFS. However, this intention is not accomplished and it seems to be coming apartmore and more. The objective of this doctoral study is recovering the language integration andprovides a unified language system for Semantic Webs.<br/>&nbsp;&nbsp;In this study, firstly, the semantics of RDF, RDFS, and OWL are investigated in commondescription based on Tarskian denotational semantics, whereas the formal way of describingsemantics in W3C Recommendations is different between RDF(S) and OWL. RDF semantics isformalized based on Tarskian denotational model theory and RDFS is extended in the same framework, but OWL semantics is mainly described in the way called Direct ModelTheoretic Semantics, which is appropriate for describing Description Logics and OWL DL. However, due to these two different styles, it has become difficult to understand both languages in the common view,and it has amounted to that OWL has become apart from RDF and RDFS. In this dissertation, an overview of RDF semantics is given in the way described in the RDF documents of W3C Recommendations. Then, OWL semantics is also investigated and formalized based in the same way as RDF, making reference to the OWL specifications in OWL Direct Model Semantics in the documents of W3C Recommendations. Since our semantic web language system is built on top of Common Lisp Object System, CLOS semantics and its computational model is also discussed. The semantic gap between OWL and object oriented languages are also pointed out. <br/>&nbsp;&nbsp;Secondly, RDF semantics is realized on top of CLOS by straightforward mapping of RDF graph,which is unidirectional labeled graph, to CLOS objects, so that a start node of edge to a CLOSobject, an edge in graph to a slotname, and an end node of edge to a slotvalue. RDFS class/instance relationship is mapped to that in CLOS, and RDFS class/superclass relationship is mapped to that in CLOS, because the semantics of RDFS is analogous to the semantics of CLOS type system. The problems arising from such straightforward mapping for RDF and RDFS are discussed and solved in the realization of our language system for SemanticWebs. Then, all OWL features are implemented on top of RDF(S) by CLOS with preserving RDF(S) semantics. We distinguish substantial sorts and nonsubstantial sorts in ontology description, and procedural subsumption computation algorithm for OWL Full is developed. The system is named SWCLOS from the acronym of Semantic Web Common Lisp Object System. <br/>&nbsp;&nbsp;Thirdly, the efficiency of SWCLOS implementation was tested by the Lehigh UniversityBenchmark (LUBM) test, and SWCLOS showed the comparable performance to other OWLreasoners, which are reported in the LUBM report. SWCLOS returned correct answers for allLUBM queries, whereas two reasoners out of three returned wrong answers for some queries in thebenchmark report. Distinctive benchmark results of SWCLOS were analyzed and improvements ofefficiency were achieved by several different engineering methods. The metamodeling capability ofSWCLOS was also demonstrated in some examples of SWCLOS metamodeling programming. <br/>&nbsp;&nbsp;Through this study, we obtained deep understanding of semantics on RDF, RDFS, and OWL,because it was necessary to realize the integration of semantic web language, namely, OWL Full, inorder to solve the semantic disparity between RDF(S) and OWL DL. For example, the subsumptionin class hierarchy is weak in RDFS but it is strong in OWL DL. The semantics of OWL DL class isakin to set theory, but the semantics of RDFS class is based on but different from set theory, ratherit is close to frame systems. The semantics of RDF(S) is basically categorized into higher orderlogic but OWL DL is notably a subset of first order logic. RDFS allows the membership loop andenables metamodeling of ontology but OWL DL cannot accept the membership loop and does notallow metamodeling. Entities in RDF universe stand in the Unique Name Assumption (UNA) forgraph nodes but entities in OWL universe does not stand in UNA for objects in ontology. RDFsemantics is not developed up to Open World but OWL semantics assumes it for WWWs. Thesehighly conceptual and technical issues must be discussed and settled in order to integrate RDF(S)and OWL. The solution for membership loop, weak/strong subsumption, and nonUnique NameAssumption, Open World Assumption are proposed and implemented in SWCLOS. <br/>&nbsp;&nbsp;In addition to these differences of semantic foundation of languages, what is worse, a misunderstanding on the interpretation of RDF semantics involved the community in the theoretical disorder against the discussion of RDF compatibility of OWL. Excessive materialization of RDF entities in OWL was coached after the argument over ‘comprehension principle’, and RDF semantics has become the focus of criticism under the pretense that ‘comprehension principle’ allows the paradox to invade upon systems. Such theoretical disorders in Semantic Web community are also discussed in order to rescue OWL Full theory from the theoretic disorder. It deserves to know that OWL 2 specifications of W3C Recommendations retracted the term ‘comprehension principle’ with no account from the documentation of W3C in the end. <br/>&nbsp;&nbsp;As its name implies, SWCLOS is not based on a logic system but based on Common Lisp Object System. It is semantically an amalgamation of CLOS and OWL on top of RDF;nevertheless it still conforms to objectoriented paradigm as programming language. It is the reason why we call it objectoriented semantic web language. The ground of enabling SWCLOS can be summarized as follows. First, the subsumption of CLOS is the same as the subsumption of RDF(S), and the structure of hierarchy and orders of CLOS classes is the same as ones of RDF(S). The dynamic property of CLOS programming and the MetaObject Protocol of CLOS enabled to tailor the semantics of language within the realms of CLOS language. In fact, it was easy to realize RDF semantics on top of CLOS, because the semantics is almost same except propertycentric or objectcentric. Then, OWL Full levelcapability is obtained in OWL by pursuing the compatibility to RDF and preserving it. <br/>

4アクセス

各種コード

  • NII論文ID(NAID)
    500000547373
  • NII著者ID(NRID)
    • 8000000549463
  • 本文言語コード
    • eng
  • NDL書誌ID
    • 023254187
  • データ提供元
    • 機関リポジトリ
    • NDL-OPAC
ページトップへ