Semantic Markup Languages: A Gentle Introduction

Transcription

1 Semantic Markup Languages: A Gentle Introduction USC/Information Sciences Institute gil@isi.edu 1

2 Outline I: The Big Picture The Semantic Web II: A Gentle Introduction XSD, RDFS, (DAML), OWL III: The Big Picture Revisited W3C s Semantic Web principles How this is changing our research in Knowledge Bases 2

3 I: THE BIG PICTURE 3

4 The Semantic Web W3C s Tim Berners-Lee: Weaving the Web : I have a dream for the Web and it has two parts. The first Web enables communication between people The Web shows how computers and networks enable the information space while getting out of the way The new Web will bring computers into the action Step 1 -- Describe: putting data on the Web in machineunderstandable form -- a Semantic Web RDF (based on XML) Master list of terms used in a document (RDF schema) Each document mixes global standards and local agreed-upon terms (namespaces) Step 2 -- Infer and reason: apply logic inference Operate on partial understanding Answering why Heuristics 4

5 Semantics and Meaning according to TBL In the extreme view, the world can be seen as only connections, nothing else. I like the idea that a piece of information is really defined only by what it s related to, and how it s related. There really is little else to meaning. The structure is everything. What matters is in the connections. It isn t the letters, it s the way they are strung together into words. [ ] into phrases. [ ] into a document. For the people, by the people: the right to link Once [ something ] was made available, it should be accesible to anyone [ ]. And it should be possible to make a link to that thing. 5

6 And There You Have I 6

7 II: THE GENTLE INTRODUCTION 7

8 The Layer Cake [TBL,XML2000] 8

9 The Layer Cake [TBL,XML2000] 9

10 URIs: Uniform Resource Identifiers (aka URLs) ftp:// The Web is an information space. URIs are the points in that space. Short strings that identify resources in the web: documents, images, downloadable files, services, electronic mailboxes, and other resources. They make resources addressable in the same simple way. They reduce the tedium of "log in to this server, then issue this magic command..." down to a single click. 10

11 11

12 Unicode A character encoding system, like ASCII, designed to help developers who want to create software applications that work in any language in the world Unicode provides a unique number for every character, no matter what the platform, no matter what the program, no matter what the language 12

13 The Layer Cake [TBL,XML2000] 13

14 Why XML (extensible Markup Language) Problems with HTML HTML design - HTML is intended for presentation of information as Web pages. - HTML contains a fixed set of markup tags. This design is not appropriate for data: - Tags don t convey meaning of the data inside the tags. - Tags are not extensible. 14

15 The Design of XML Tags can be used to represent the meaning of data/information separates syntax (structural representation) from semantics => only syntax is considered in XML There is no fixed set of markup tags - new tags can be defined Underlying data model is a tree structure XML is the new ASCII -- Tim Bray 15

16 Simple XML Example <Bookstore> <Book ID= 101 > <Author>John Doe</Author> <Title>Introduction to XM L</Title> <Date>12 June 2001</Date> <ISBN> </ISBN> <Publisher>XYZ</Publisher> </Book> <Book ID= 102 > <Author>Foo Bar</Author> <Title>Introduction to XSL</Title> <Date>12 June 2001</Date> <ISBN> </ISBN> <Publisher>ABC</Publisher> </Book> </Bookstore> Make up your own tags Sub-elements XML by itself is just hierarchically structured text 16

17 17

18 XSD: XML Schema Definition Written in the same syntax as XML documents (unlike XML DTDs!) Elements and attributes Enhanced set of primitve datatypes. Wide range of primitive data types, supporting those found in databases (string, boolean, decimal, integer, date, etc.) Can create your own datatypes (complextype) - Can derive new type definitions on the basis of old ones (refinement) Can have constraints on attributes Examples: maxlength, precision, enumeration, maxinclusive (upper bound), mininclusive (lower bound), etc. 18

19 XSD (XML Schema) Example <?xml version="1.0"?> <xsd:schema xmlns:xsd= LSchema targetnamespace=" xmlns= > <xsd:element name="bookstore"> <xsd:complextype> <xsd:sequence> <xsd:element ref="book" minoccurs="1" maxoccurs="unbounded"/> </xsd:sequence> </xsd:complextype> </xsd:element> <xsd:element name="book"> <xsd:complextype> <xsd:sequence> <xsd:element ref="title" minoccurs="1" maxoccurs="1"/> <xsd:element ref="author" minoccurs="1" maxoccurs= unbounded /> <xsd:element ref="date" minoccurs="1" maxoccurs="1"/> <xsd:element ref="isbn" minoccurs="1" maxoccurs="1"/> <xsd:element ref="publisher" minoccurs="1" maxoccurs="1"/> </xsd:sequence> </xsd:complextype> </xsd:element> <xsd:element name="title" type="xsd:string"/> <xsd:element name="author" type="xsd:string"/> <xsd:element name="date" type="xsd:date"/> <xsd:element name="isbn" type="xsd:integer"/> <xsd:element name="publisher" type="xsd:string"/> </xsd:schema> Prefix xsd refers to the XMLSchema namespace xmlns refers to the default namespace Defining the element Bookstore as a complex Type Containing a sequence of 1 or more Book elements When referring to another Element, use ref The Author can be 1 or more Element definitions Notice the use of more meaningful data types 19

20 XSL [XML Stylesheet Language] 20

21 XML: Tools/Software XML Spy By far, the most comprehensive editor. Handles XML files, DTD s, XSL files, as well as XSD (XML Schema). Unfortunately only a 30 day trial version. XML Notepad Microsoft XML Notepad is a simple application for building and editing small sets of XML-based data. Freeware. XML Pro XML Pro is a top-notch XML editor but it doesn t include as many features as XML Spy. Shareware. You can also validate your XML files by just opening them with IE5.0 or above. It checks if the XML file is well-formed or not, and also validates against a DTD (if specified on the DOCTYPE declaration Some nice & short Tutorials on XML/XSL/DTD/XML Schemas can be found at: 21

22 Summary of the XML+ NS +XSD Layer The Power of Simplicity When I designed HTML, I chose to avoid giving it more power than it absolutely needed a principle of least power, which I have stuck to ever since. I could have used a language like Knuth s Tex but - - TBL Keeps the principles of SGML in place but its spec is thin enough to wave ϑ To say you are Using XML is sort of like saying you are using ASCII Using XSD (XML Schema) makes a lot more sense 22

23 The Layer Cake [TBL,XML2000] 23

24 Where XML & XML Schemas Fail No semantics! <book> <title> </title> <author> </author> <isbn> </isbn> </book> <bookstore> <book> </book> <mgzine> </mgzine> </bookstore> Will XML scale in the metadata world? 1. The order in which elements appear in an XML document is often meaningful. This seems highly unnatural in the metadata world. Furthermore, maintaining the correct order of millions of data items is impractical. 2. XML allows constructions that mix up some text along with child elements, which are hard to handle. Ex. <topelem>this is some character string data <elem> this is a child <subelem>this is another child</subelem> </elem> </topelem> 24

25 RDF (Resource Description Framework) RDF provides a way of describing resources via metadata (data about data) It restricts the description of resources to triplets (subject,predicate,object) It provides interoperability between applications that exchange machine understandable information on the Web. The original broad goal of RDF was to define a mechanism for describing resources that makes no assumptions about a particular application domain, nor defines (a priori) the semantics of any application domain. Uses XML as the interchange syntax. Provides a lightweight ontology system. The formal specification of RDF is available at: 25

26 RDF Syntax Subject, Predicate and Object Triplets (Tuples) Subject: The resource being described. Predicate: A property of the resource Object: The value of the property A combination of them is said to be a Statement (or a rule) Author John Doe A web page being described A property of the web page (author) The value of the predicate (here the author) [Subject] [Predicate] [Object] 26

27 RDF Example <?xml version="1.0"?> <rdf:rdf xmlns:rdf=" xmlns:s=" <rdf:description about=" Subject <s:author>john Doe</s:Author> Author (property of the subject) </rdf:description> (Also a resource) </rdf:rdf> Object. Can also point to a resource The above statement says : The Author of is John Doe Namespace for the RDF spec Namespace s, a custom namespace In this way, we can have different objects (resources) pointing to other objects (resources), thus forming a DLG (Directed Line Graph) You can also make statements about statements reification Ex: xyz says that The Author of is John Doe 27

28 RDF Schema A schema defines the terms that will be used in the RDF statements and gives specific meanings to them. Example: <rdf:rdf xml:lang="en" xmlns:rdf=" xmlns:rdfs=" <rdf:description ID="M otorvehicle"> <rdf:type resource=" <rdfs:subclassof rdf:resource=" </rdf:description> <rdf:description ID="PassengerVehicle"> <rdf:type resource=" <rdfs:subclassof rdf:resource="#m otorvehicle"/> </rdf:description> <rdf:description ID="Truck"> <rdf:type resource=" <rdfs:subclassof rdf:resource="#m otorvehicle"/> </rdf:description> RDF Schema Namespace An ID attribute actually defines a new resource Resource is the top level class PassengerVehicle is a subclass of MotorVehicle 28

29 Example (cont..) <rdf:description ID="Van"> <rdf:type resource=" <rdfs:subclassof rdf:resource="#m otorvehicle"/> </rdf:description> <rdf:description ID="M inivan"> <rdf:type resource=" <rdfs:subclassof rdf:resource="#van"/> <rdfs:subclassof rdf:resource="#passengervehicle"/> </rdf:description> <rdf:description ID="registeredTo"> <rdf:type resource=" <rdfs:domain rdf:resource="#m otorvehicle"/> <rdfs:range rdf:resource="#person"/> </rdf:description> Multiple Inheritance Domain of a property <rdf:description ID="rearSeatLegRoom"> <rdf:type resource=" <rdfs:domain rdf:resource="#passengervehicle"/> <rdfs:domain rdf:resource="#m inivan"/> <rdfs:range rdf:resource=" </rdf:description> </rdf:rdf> Range of a property 29

30 RDF: Tools/Resources SirPAC A Simple RDF Parser & Compiler. It parses the RDF, and validates it. It also generates the tuples and even draws a graph of the data model. FRODO A visualization tool for RDF schema representations. Reggie A Nice Metadata Editor. Java based simple user interface to describe a web resource. Can mail the metadata file to yourself after finished editing. Protégé Editor of ontologies in practically any language you care about. Open source. 30

31 Summary: RDF & RDF Schema layer Minimalist model - (thing), Class, Property Subproperty, Subclass Domain & Range RDF Schema: a W3C recommendation Feb 04 Efficient storage and retrieval Triple store using database backends 31

32 The Layer Cake [TBL,XML2000] 32

33 Limitations of RDF - Cannot define properties of properties (unique, transitive) - No equivalence, disjointness, etc. - No mechanism of specifying necessary and sufficient conditions for class membership. Example: If it is given that XYZ has a car which is 7ft high, has wide wheels and loading space is 4 cub.m, then we should be able to reason that XYZ has an SUV, as given by the necessary and sufficient conditions for being an SUV : height > 4ft & wide wheels & loading space > 2 cub.m 33

34 DAML+OIL s History W3C s Semantic Web Activity: - RDF and metadata markup efforts to represent data in a machine understandable form. DARPA started the DARPA Agent Markup Language (DAML) program. possibly with ARPANET -> Internet in mind EC (European Commission) funding programs - Ontology Interchange Language (OIL) - logic based language. - brings logic and inference to the Semantic Web DAML+OIL: 34

35 DAML+OIL ( It builds on earlier W3C standards such as RDF and RDF Schema. DAML extends RDF and RDFS with richer modelling primitives. disjointwith, intersectionof, oneof, cardinality Able to provide properties of properties uniqueness, transitivity, etc. Current version DAML+OIL provides a semantic interpretation (model-theoretic semantics) 35

36 DAML+OIL substrate: Description Logics Classes are defined in terms of other classes/relations Relations are first class citizens Powerful inference algorithms: Subsumption: is classa a subclass of classb given their definitions? Recognition: is instancea of classa? Classification: automatic reorganization of class hierarchy based on definitions of classes 36

37 An Example (from Start of an ontology (about = implies this document) The label is not used for logical interpretation Can explicitly specify the set of Females to be disjoint with the set of Males The Person class is defined later To be read conjunctively. A man is a sub-class of Person and a Male 37

38 Example (contd..) An objectproperty relates objects to objects Describes the element which encloses this Property Describes the value of the Property Note: Contrary to RDF, DAM L takes the intersection of the domains/ranges if multiple domains/ranges are specified A datatype property relates an object to a primitive datatype value The XM L Schema datatype is referenced here USC INFORMATION SCIENCES INSTITUTE The Restriction defines an anonymous class of all things that satisfy the restriction. Restrictions on the property hasparent (only for the Person class Local scope, as opposed to rdfs:range) A person can have only another Person as it s parent A Person can have only 1 Father 38

39 Example (contd..) Restrictions on the property hasparent An animal can have exactly 2 parents Restrictions on the property hasspouse A p Further constructs that the example doesn t use : Properties: TransitiveProperty (hasancestor), UniqueProperty (hasmother), inverseof(haschild -> hasparent), etc. Classes: intersectionof (a daml:collection), unionof (a daml:collection), sameclassas, complementof, etc. 39

40 DAML+OIL Most used knowledge representation language in history : 5M statements on 20K web pages 13M triples RDF-based average 20k hits/day Submit a query, many things are there: 40

41 DAML References/Tools DAML Viewer: It provides a means to view the instances found in a DAML document. DAML Crawler Results: A list of.daml files on the internet A DAML Validator A DAML example explained: It has the same example as in the slides, discussed in detail. 41

42 OWL: Web Ontology Language Takes DAML+OIL as input, follows standard W3C process OWL: Exte 42

43 OWL features RDF schema (class, subclass, property, subproperty Restrictions Range, domain Local, global Existential Cardinality Combinators Union, intersection Complement Symmetry Transitivity Mappings Equivalent inverse 43

44 OWL Lite A restricted version of OWL, designed to support efficient reasoning All of OWL except: owl:oneof owl:unionof owl:complementof owl:hasvalue owl:disjointwith owl:datarange and a few other restrictions Increasing expressivity is provided in OWL-DL and OWL-Full 44

45 The Layer Cake [TBL,XML2000] 45

46 To Learn More About These Languages... more detailed tutorial slides l 46

47 The Layer Cake DAML+OIL DAML-O OIL NOTATION 3 HTML XHTML RDFS XQUERY XTM RDF XML 47

48 The Layer Cake DAML+OIL HTML XHTML DAML-O RDFS XQUERY XTM RDF XML OIL OWL NOTATION 3 48

49 III: THE BIG PICTURE REVISITED 49

50 W3C s Semantic Web Principles 1. Everything identifiable is in the Semantic Web (URIs!) 2. Partial information Anyone can say anything about anything 3. Web of trust All statements on the Web occur in some context 4. Evolution Allow combining independent work done by different communities 5. Minimalist design Make the simple things simple, and the complex things possible Standardize no more than is necessary 50

51 Hypertext: Then and Now SOTA circa 1990: Dynatext s electronic book A book had to be compiled (like a program) in order to be displayed efficiently A central link database, to make sure there were no broken links Text that was fixed and consistent (a whole book) WWW: Links can be added and used at any time Distributed (must live with broken links!) Decentralized 51

52 Knowledge Representation: Now and Tomorrow To webize KR in general is, in many ways, the same as to webize hypertext. Replace identifiers with URIs. Remove any requirement for global consistency. Put any significant effort into getting critical mass. Sit back. -- TBL 52

53 Other Topics Semantic web services DAML-S, OWL-S Query and Rule languages Web of trust 53

54 TRELLIS [Gil & Ratnakar, ISWC 02]: Developing the Web of Trust One Citizen at a Time Capture decisions and opinions as the user finds, analyzes, and uses information (Web of Trust) annotate the reasons, judgment, and purpose that make information meaningful keep track of contradictory, related, and rejected information annotate derivation of conclusions and formulation of hypotheses add structure and formalization incrementally semantic markup language Derive an assessment of info sources based on individual opinions Try it out at 54

55 Editors for Manual Annotation of Content 55

56 Automatic Annotation of Documents 56

57 Editors of Ontologies and Schemas 57

58 Browsers of Semantic Annotations 58

59 CS Aktive Space (challenge.semanticweb.org) 59