1、第第16章章語義語義WEB與與WEB數(shù)據(jù)庫數(shù)據(jù)庫 Semantics Web RDF and RDFs Web Ontologies(OWL) Description Logic Web DatabaseWeb歷史歷史 歷史歷史 1990年，年，Tim Berners-Lee發(fā)明發(fā)明Web 第一代第一代(Web 1.0) ：手寫的：手寫的HTML/Web頁頁 第二代第二代(Web 1.5)：動(dòng)態(tài)：動(dòng)態(tài)Web頁，但還是頁，但還是HTML 第三代第三代(Web 2.0)：Semantic WebWeb現(xiàn)狀現(xiàn)狀 資源通過語義有限的資源通過語義有限的鏈接相互關(guān)聯(lián)鏈接相互關(guān)聯(lián) 一個(gè)系統(tǒng)無法自動(dòng)判一個(gè)系統(tǒng)

2、無法自動(dòng)判定定/理解其它系統(tǒng)中的理解其它系統(tǒng)中的資源的含義資源的含義 如：圖中資源如：圖中資源D通過通過HREF與資源與資源A和和C鏈鏈接，接， 但是無法知道但是無法知道A和和C的含義的含義CABD問題何在？問題何在？ HTML&URL：只包含如何顯示某只包含如何顯示某條信息的標(biāo)記，如條信息的標(biāo)記，如, 表示什表示什么顏色，什么字體么顏色，什么字體等等內(nèi)容很容易為人所內(nèi)容很容易為人所理解，但是對(duì)計(jì)算理解，但是對(duì)計(jì)算機(jī)來說卻不是那么機(jī)來說卻不是那么容易容易Web互操作性互操作性 Web通過已建立的標(biāo)準(zhǔn)保證在各層次上的通過已建立的標(biāo)準(zhǔn)保證在各層次上的互操作性互操作性 TCP/IP協(xié)議讓人們

3、無需關(guān)心如何傳輸二進(jìn)協(xié)議讓人們無需關(guān)心如何傳輸二進(jìn)制數(shù)據(jù)制數(shù)據(jù) HTTP和和HTML提供了超文本的傳輸和表現(xiàn)提供了超文本的傳輸和表現(xiàn)的標(biāo)準(zhǔn)的標(biāo)準(zhǔn) XML實(shí)現(xiàn)了語法層的互操作實(shí)現(xiàn)了語法層的互操作Web互操作性互操作性 Web還需要更高層次上的互操作性標(biāo)準(zhǔn)還需要更高層次上的互操作性標(biāo)準(zhǔn) 語義互操作：語義互操作：RDF，RDF Schema，OWL什么是語義什么是語義WEB？ 語義語義WEB是基于是基于Web的機(jī)器可理解的信息網(wǎng)絡(luò)，的機(jī)器可理解的信息網(wǎng)絡(luò)，是一個(gè)覆蓋全球的數(shù)據(jù)網(wǎng)或數(shù)據(jù)庫是一個(gè)覆蓋全球的數(shù)據(jù)網(wǎng)或數(shù)據(jù)庫 原則原則 一切可以確定的內(nèi)容（人，時(shí)間，事件，物體，一切可以確定的內(nèi)容（人，時(shí)間，事

4、件，物體，事物，等）都包括在一個(gè)基于知識(shí)的網(wǎng)絡(luò)中事物，等）都包括在一個(gè)基于知識(shí)的網(wǎng)絡(luò)中 每個(gè)實(shí)體（每個(gè)實(shí)體（entity)都有一個(gè)統(tǒng)一資源標(biāo)識(shí)都有一個(gè)統(tǒng)一資源標(biāo)識(shí)(URI) 信息不一定是完整的，信息不一定是完整的，Web沒有盡頭，語義網(wǎng)也沒有盡頭，語義網(wǎng)也是如此，任何找到的信息都只是部分信息是如此，任何找到的信息都只是部分信息 網(wǎng)絡(luò)需要不斷進(jìn)化和完善網(wǎng)絡(luò)需要不斷進(jìn)化和完善 使簡(jiǎn)單事物更加簡(jiǎn)單，使復(fù)雜事物的盡可能簡(jiǎn)單使簡(jiǎn)單事物更加簡(jiǎn)單，使復(fù)雜事物的盡可能簡(jiǎn)單 從簡(jiǎn)單應(yīng)用開始，逐漸增加復(fù)雜應(yīng)用從簡(jiǎn)單應(yīng)用開始，逐漸增加復(fù)雜應(yīng)用Semantic Web的目標(biāo)的目標(biāo) Web上定義和鏈接的數(shù)據(jù)不僅能顯示，

5、而上定義和鏈接的數(shù)據(jù)不僅能顯示，而且可以被機(jī)器自動(dòng)地處理、集成和重用且可以被機(jī)器自動(dòng)地處理、集成和重用 只有當(dāng)數(shù)據(jù)不僅可以被人而且可以被機(jī)器只有當(dāng)數(shù)據(jù)不僅可以被人而且可以被機(jī)器自動(dòng)地共享和處理的時(shí)候，自動(dòng)地共享和處理的時(shí)候，Web的潛力才的潛力才可能得到充分發(fā)揮可能得到充分發(fā)揮 機(jī)器可理解機(jī)器可理解并不意味著如機(jī)器能夠理解人并不意味著如機(jī)器能夠理解人類的語言類的語言 機(jī)器機(jī)器”可理解可理解”是指：是指：機(jī)器能夠根據(jù)明確機(jī)器能夠根據(jù)明確定義的語義，通過執(zhí)行明確定義的操作，定義的語義，通過執(zhí)行明確定義的操作，解決明確定義的問題解決明確定義的問題語義語義Web體系結(jié)構(gòu)體系結(jié)構(gòu)語義語義Web體系結(jié)構(gòu)體

6、系結(jié)構(gòu)名稱名稱描述描述1Unicode+URIUnicode處理資源編碼，處理資源編碼，URI處理資源標(biāo)識(shí)處理資源標(biāo)識(shí)2XML+NS+xmlschemaXML處理信息內(nèi)容表示，處理信息內(nèi)容表示，NS處理命名空間，處理命名空間，xmlschema規(guī)定文檔結(jié)構(gòu)規(guī)定文檔結(jié)構(gòu)3RDF+rdfschemaRDF元數(shù)據(jù)表示方法，元數(shù)據(jù)表示方法，rdfschema資源類別（詞匯表）資源類別（詞匯表）表示方法。本層是語義的最簡(jiǎn)模型，具有廣泛的互操表示方法。本層是語義的最簡(jiǎn)模型，具有廣泛的互操作性作性4Ontology Vocabulary共享的本體詞匯表。包含更多的元信息共享的本體詞匯表。包含更多的元信息（m

7、etainformation）,如傳遞性、唯一性、明確性；如傳遞性、唯一性、明確性；產(chǎn)生附加應(yīng)用；廣泛的互操作性產(chǎn)生附加應(yīng)用；廣泛的互操作性5Logic業(yè)務(wù)業(yè)務(wù)單調(diào)邏輯的通用語言；多個(gè)推理機(jī)（可能推單調(diào)邏輯的通用語言；多個(gè)推理機(jī)（可能推理能力不同）；基于規(guī)則的系統(tǒng)可以向語義理能力不同）；基于規(guī)則的系統(tǒng)可以向語義Web輸出輸出信息，但不能從中獲得信息信息，但不能從中獲得信息6Proof證據(jù)確認(rèn)（證據(jù)確認(rèn)（Validate Proofs）7Trust可信可信RDF and RDFsRDF Basic Resource Description Framework (RDF) is a foundat

8、ion for processing Web metadata It provides interoperability between applications that exchange machine-understandable information on the WebRDF Basic What it doesIntegrates a variety of applications from library catalogs and world-wide directories to syndication and aggregation of news, software, a

9、nd content to personal collections of music, photos, and eventsRDF Basic How it looks likeXML syntax-based How it worksThe RDF specifications provide a lightweight ontology system to support the exchange of knowledge on the WebRDF Basic Self Describing : a framework for describing resources Resource

10、s: things that can be named with URIs (http, urn, .) Description: statements about the properties of these resources Framework: a common model for diversityRDF BasicBuilds onURI: Uniform Resource IdentifiersXML: Extensible Markup LanguageRDF Basic The syntax of RDF may uses the Extensible Markup Lan

11、guage XML RDF and XML are complementary: RDF is a model of XML metadata RDF Encoding, transportation, file storage (such as internationalization, character sets, etc.) relies on the support of XMLURIThe Semantic Web will build on the distributed nature of the Web:No central naming authoritySchema de

12、finitions:Not implied in applications (cf. Web browsers and HTML DTD)Accessible in a machine-understandable format using a URIURI URI=URL+URN URI Examples /mm6/my.xml#root().child(1) mailto:urn:isbn:0764532367（某本書）（某本書）RDF - Architectu

13、ral Aims W3C Metadata needs:Defines a convention for applications that exchange metadata on the WebXML used for serialization syntaxVocabulary semantics defined by communities of expertise, not W3CFine-grained mixing of diverse metadata (NS)URI specification for Web identifiersRDF for Semantic Annot

14、ation(標(biāo)注標(biāo)注) RDF provides metadata about Web resourcesObject - Attribute- Value triples It has an XML syntax Chained triples form a graphhttp:/www.cs.vu.nl/jbroeks/site-ownerJeen4417876telW3Cexplains/RDFsite-owner 4417876RDF Basic ModelSubjectObjectPredicate Resources: All things bein

15、g described by RDF expressions are called resources Properties: A property is a specific aspect, characteristic, attribute, or relation used to describe a resource. It is a subset of Resources Statements: A specific resource together with a named property plus the value of that property for that res

16、ource is an RDF statement Subject, Predicate, ObjectRDF Basic Model/rdf.html TitleAn Introductionto RDFTitle An Introduction to RDF Serialization syntax based on XML Understandable and processable by machinesRDF Basic ModelR1R2Nodes are resources connected by named propertiesP1R1“

17、foo”The degenerate(退化) case is an arc terminating in a fixed valueP1R1R2R3R6R4An RDF description consists of a directed graph of arbitrary complexityR7R5R8P1P2P3P4P5P6P7RDF Vs XMLRDF:An XML application“Not just tags” RDF makes use of a formal modelResourceValuePropertyTypePropertyRDF Data ModelBrian

18、written by05-Mar-02onpage.htmlResourcehas property valuePage.html written-byBrianKnown as triples or tuplesRDF Primitives RDF NameSpacexmlns:rdf=“/1999/02/22-rdf-syntax-ns#”rdf:Propertyrdf:typerdf:valueStatement about Statementrdf:Statement, rdf:predicate, rdf:subject, rdf:objectrdf:

19、Bag, rdf:Seq, rdf:Alt 輔助輔助(未明確定義未明確定義)：rdf:rdf, rdf:ID, rdf:about, rdf:nodeID, rdf:Description, rdf:parseTypeRDF Collection Containers Multiple occurrences of the same PropertyType doesnt establish a relation between the values Eg. The Millers own a boat, a bike, and a TV set RDF defines three speci

20、al Resources:BagSequenceAlternativeRdf:Bag/Students/Amy/Students/Tim/Students/John/Students/Mary/Students/Suebagid1/courses/6.001studentsrdf:typerdf:_1rdf:_2rdf:_3rdf:_4rdf:_5The students incourse 6.001 are Amy, Tim,John, Mary, and SueRepeated Property Consider as an example the relationship between

21、 a writer and her publications, We might have the sentence: Sue has written “Unix Operating System, “Introduction to Java, “O-O Programming. /people/Sue/Book/Unix./Book/Java./Book/O-O.publicationpublicationpublicationReification(具體化具體化) If we write the sentenceRalph Swick says that Ora Lassila is th

22、e creator of the resource /Home/Lassila. To model the example above, we could attach another property to the reified statement (say, attributedTo) with an appropriate value (in this case, Ralph Swick) Using base-level RDF/XML syntax, this could be written as:rdf:Statementhttp:/www.w3

23、.org/Home/Lassilas:Creatora:sayrdf:typerdf:subjectrdf:objectrdf:predicateOra LassilaRalph SwickStatements about statements Transform them into Resources Ralph Swick believes that the creator of the resource /Home/Lassila is Ora Lassilardf:Statementrdf:type genid1Ralph Swickb:believed

24、By/Home/Lassilardf:subjectOra Lassilardf:objects:Creatorrdf:predicates:CreatorNote that this shorthand example includes additional facts in the model not represented by the example in Figure This shorthand usage expresses facts about Ralphs statements and also facts about Oras home p

25、ageBag and Statement-about-statementStatement about statementProperty with Structured Value The individual referred to by employee id85740 is named Ora Lassila and has the email address The resource /home/lassila was created by this individual/home/lassi

26、laOra Lassilacreatoremployee-id/85740nameemail Subject (Resource) /Home/Lassila Predicate (Property) Creator Object (literal) Ora Lassila/Home/LassilaOra LassilaCreatorFigure 1: Simple node and arc diagramOra Lassila is the creator of the resource http:/w

27、/Home/LassilaRDF Example Ora Lassila is the creator of the resource /Home/Lassila Ora Lassila xmlns:s=/schema/ The root element of an RDF document is RDF.Each Description element indicate a Subject.RDF Example(12/15) The individual whose name is Ora Lass

28、ila, email , is the creator of /Home/Lassila/Home/LassilaOra LCreatorNameEmailRDF ExampleRDF Data Model (A kind of) Directed Labeled Graphedges represent propertiesnodes representresources (ellipses)literals (rectangles) are sharedthere may be multi

29、ple edges between two nodesthere may be multiple edges with the same label pointing to different nodesthe graph can contain cycles the graph is not necessarily connected28Pic1Person1filmFrameagecreatedBydepictingPerson2depictingsubjectobjectpredicateRDF Data Model as Triplets Definition: An RDF mode

30、l M is a finite set of triplets (also called statements) of the form:M R x U x (R L) Where U is the set of (references to)resource identifiers or URIsL is the set of literals (string like elements) which denote the actual content (data)R= U B is the set of all resources (both blank and labeled) B is

31、 the set of blank nodes (nodes that dont have a label)subjectobjectpredicateRDF Data Model as TripletsExample: Pic1, filmFrame, “28” Pic1, depicting, Person1 Pic1, createdBy, Person1 Pic1, depicting, Person2 Person1, age, “28”28Pic1Person1filmFrameagecreatedBydepictingPerson2depictingRDF Data Model

32、as Triplets Implications: properties have always labels (URIs), i.e., there are no “blank” propertiesliteral nodes cannot have properties (they can stand only as property values)it is possible to make statements about properties, but beware a statement about one property propagates傳播to all propertie

33、s with the same label in the modelIn fact all properties with the same label are treated as one resourceRDF Data Model as TripletsExample: Pic1, filmFrame, “28” Pic1, depicting, Person1 Pic1, createdBy, Person1 Pic1, depicting, Person2 Person1, age, “28”28Pic1Person1filmFrameagecreatedBydepictingPer

34、son2depictingndepicting, how, “nicely”depictingnicelyhownicelyhowhowAny ideas?RDF Data Model as TripletsSolution: Pic1, depicting, b1 b1, how, “nicely” b1, whom, Person1 Pic1, createdBy, Person1 Pic1, depicting, b2 b2, whom, Person2Pic1Person1createdBydepictingPerson2depictingb1b2whomwhomnicelyhowRD

35、F Schema A modeling language on top of RDF (expressed in RDF) Introduces the following modeling primitives:rdf:Propetyrdfs:Classrdf:typerdf:subClassOfrdf:subPropertyOfrdfs:Literalrdf:rangerdf:domainrdfs:Resourcerdfs:subClassOfrdf:typeRDF Schema: Classes Resources may be divided into groups called cl

36、asses DefinitionThe extent of a class is the set of all resources that are linked to that class by the rdf:type propertyTwo classes with the same extent are not necessarily the same, i.e. they might have different propertiesClasses can be subsumed包容by other classes with the transitive rdfs:subClassO

37、f propertyRDF Schema: Properties Properties are linked to classes by specifying their rdfs:domain and rdfs:range,. i.e., they are defined externally wrt classes (they are first-class citizens) Definition The set Domain(Range) of a property p is defined as the intersection of extents of all classes c

38、 indicated by p,rdfs:domain (rdfs:range),cThe extent of a property p is the set E of all triplets x, p, y, where xDomain and y RangeIf a property has more than one rdfs:domain (rdfs:range), objects(subjects) using this property are instances of all classes stated in the rdfs:domain (rdfs:range) .pro

39、perties can be (also) subsumed with the transitive rdfs:subPropertyOf propertyRDF Schema: ExampleDefinition uses the Data model of RDFDefined in terms ofRDFRDF SchemaOntologyLanguageInstance DataIs extension ofRDF(S) ExtensibilitynDefine an Ontology of your Language with RDF Schema (like RDF-Schema

40、itself)nDescribe Instance Data using your new VocabularynAdvantage: all Languages use the same Data Model (simplifies Interoperability) RDF has “Non-standard” semantics Semantics given by RDF Model Theory (MT) In RDF MT, an interpretation I of a vocabulary V consists of: IR: a non-empty set of resou

41、rces (corresponds to )IS: a mapping from V into IR (corresponds to I )IP: a distinguished subset of IR (the properties)A vocabulary element v V is a property iff IS(v) IPIEXT: a mapping from IP into the powerset of IRXIRI.e., property elements mapped to subsets of IRXIRIL:a mapping from typed litera

42、ls into IRRDF/RDF(S) SemanticsRDF/RDF(S) Semantics Class interpretation ICEXT induced by IEXT(IS(type)ICEXT(C) = x | (x,C) IEXT(IS(type) RDF(S) adds constraints on models(x,y), (y,z) IEXT(IS(subClassOf) ) = (x,z) IEXT(IS(subClassOf)RDFS simply adds semantic conditions and axiomatic triples that give

43、 meaning to schema vocabularyClass interpretation ICEXT simply induced by rdf:type, i.e.:x is in ICEXT(y) if and only if is in IEXT(IS(rdf:type)Other semantic conditions include:If is in IEXT(IS(rdfs:domain) and is in IEXT(x) then u is in ICEXT(y)If is in IEXT(IS(rdfs:subClassOf) then x and y are in

44、 IC and ICEXT(x) is a subset of ICEXT(y)IEXT(IS(rdfs:subClassOf) is transitive and reflexive on ICAxiomatic triples include:rdf:type rdfs:domain rdfs:Resource rdfs:domain rdfs:domain rdf:PropertyRDF/RDF(S) Semantic ConditionsRDF/RDF(S) Semantics InterpretationRDF/RDF(S) Semantics Interpretation Exam

45、ple RDF Schema:RDFS-Closure Definition A model M is RDFS-closed if the following holds: 1. x,y,z M, y, rdf:type, rdf:Property M2. x,y,z, y, rdfs:domain,u M, x, rdf:type, u M3. x,y,z, y, rdfs:range,u M, z, rdf:type, u M4. x,y,z M, x, rdf:type, rdfs:Resource M5. x,y,z M where z U, z, rdf:type, rdfs:Re

46、source M6. x,rdfs:subPropertyOf,y and y,rdfs:subPropertyOf,z M x,rdfs:subPropertyOf,z M7. x,y,z and y,rdfs:subPropertyOf,u M x,y z M8. x, xrdf:type, rdfs:Class M x,rdfs:subClassOf, rdfs:Resource M9. x,rdfs:subClassOf,y and y,rdfs:subClassOf,z M x,rdfs:subClassOf,z M10. x,rdfs:subClassOf,y and a,rdf:

47、type,x M a,rdf:type,y M These can be also considered as inference rules, i.e. how to derive an RDFS-closed graphAside: Semantics and Model Theories Ontology/KR languages aim to model (part of) world Terms in language correspond to entities in world Meaning given by, e.g.:Mapping to another formalism

48、, such as (First Order Logic) FOL, with own well defined semanticsor a Model Theory (MT) MT defines relationship between syntax and interpretationsCan be many interpretations (models) of one piece of syntaxModels supposed to be analogue of (part of) worldE.g., elements of model correspond to objects

49、 in worldFormal relationship between syntax and modelsStructure of models reflect relationships specified in syntaxInference (e.g., subsumption) defined in terms of MTAside: Set Based Model TheoryMany logics (including standard First Order Logic) use a model theory based on Zermelo-Frankel set theor

50、yThe domain of discourse (i.e., the part of the world being modelled) is represented as a set (often refered as )Objects in the world are interpreted as elements of Classes/concepts (unary predicates) are subsets of Properties/roles (binary predicates) are subsets of X (i.e., 2)Ternary predicates ar

51、e subsets of 3 etc.The sub-class relationship between classes can be interpreted as set inclusionDoesnt work for RDF, because in RDF a class (set) can be a member (element) of another class (set)In Z-F set theory, elements of classes are atomic (no structure)Aside: Set Based Model Theory An interpre

52、tation I is a tuple （, I ） is the domain (a set)I is a mapping that mapsNames of objects to elements of Names of unary predicates (classes/concepts) to subsets of Names of binary predicates (properties/roles) to subsets of X And so on for higher arity predicates (if any)Aside: Set Based Model Theory

53、 Example Formally, the vocabulary is the set of names we use in our model of (part of) the worldDaisy, Cow, Animal, Mary, Person, Z123ABC, Car, drives, Aside: Set Based Model Theory ExampleWorldInterpretationDaisy isA CowCow kindOf AnimalMary isA PersonPerson kindOf AnimalZ123ABC isA Carha,bi, abMod

54、elMary drives Z123ABCFOL ThesisBase languages on established FO hierarchyPropositional logicDecidable FOL subsets (e.g., DL, Horn)Undecidable FOL subsets Full FOL (and even HOL)Higher layers extend syntaxUpwards compatibility, i.e., syntax retains same meaning in higher layersSemantics via FOL mappi

55、ng or standard FO model theoryIndividual i ! element of domain (iI 2 D)Class C ! sets of elements (CI D)Property P ! binary rel on D (PI D D)(Dis)advantages of FOL ThesisProsBased on well known and extensively studied formalismWealth of theoretical knowledge and practical experienceFamily of sub-lan

56、guages with well known formal propertiesE.g., decidability, complexityHighly optimised reasoners for FOL and many sub-languagesE.g., DL reasoners, Horn (rule) reasoners, FOL proversMapping to FOL provides easy integration, e.g., of DL and Horn languagesFO subset of RDFS fits well in this frameworkCo

57、nsNo classes as instances (unless extended to HOL)Relatively poor fit with full RDFSCan be axiomatised in FOL, but may damage semantic interoperability and computational propertiesAxiomatisation An Axiomatisation can be used to embed RDFS in FOL, e.g.:Triple x P y translated as holds2(P,x,y)Axioms c

58、apture semantics of language, e.g.: Axiomatisation Problems with axiomatisations includeMay require large and complex set of axiomsDifficult to prove semantics have been correctly capturedAxiomatisation may greatly increase computational complexityRDFS ! undecidable (subset of) FOLNo interoperabilit

59、y unless all languages similarly axiomatisedE.g., in DAML+OIL, C subClassOf D equivalent to 8 x.C(x) ! D(x)But have to axiomatise as holds2(subClass, C, D)RDF Thesis(12/7)All languages based on triplesTriple based syntaxSemantics compatible with semantics of triples as defined by RDF MTUpwards &

60、 downwards compatibilitySyntax retains same meaning in higher layersHigher layer syntax is valid in lower layersSemantics via RDF model theorySimilar to CL, but only binary predicatesLanguage syntax also in domain of discourseHigher layers impose additional constraints on modelsSyntax must be encoded as tri