1. Ontologies An introduction and overview

2. © Per Flensburg 2 Who am I? Per Flensburg, Professor in Informatics Started VXU 1996 Before that Copenhagen Business School 1986- 1996 Before that Lund University 1972-1986 Interest up to 1999: Participative Design 1999- : Ontologies Research unit: 2000

3. © Per Flensburg 3 Basic concepts from my point of view Data, information, content and knowledge Ontology, semantics

4. © Per Flensburg 4 Data 42, Gothenburg, capitol of Sweden Symbol strings without meaning: Words without meaning:

5. © Per Flensburg 5 Information Data put into a (syntactical) structure Ex: The capitol of Sweden is Stockholm Information describes a relation between certain entities It does not need to correspond to reality. Replacing “Stockholm” by “Gothenburg” does not change the status as information. If the information corresponds to reality it is called a fact. The structure does not necessary need to be a grammatical one.

6. © Per Flensburg 6 A table StubbheadConeswinger0407072 GrimsfeldCrwth0407062 TurbinTravers0406063 It is information, but it has no meaning at all!

7. © Per Flensburg 7 Something is added.... CustomerOrdered partDay of orderQuantity StubbheadConeswinger0407072 GrimsfeldCrwth0407062 TurbinTravers0406063 We add metadata, thus providing some context in which the information can be interpreted and given meaning. This we call content

8. © Per Flensburg 8 Content Information provided with some metadata In text or speech this metadata is implicit in the grammatical structure Metadata is simply an explicit description of the structure of the information Metadata gives a hint about the meaning of the information. It is sometimes called semantics. However, I prefer using semantics as description of metadata

9. © Per Flensburg 9 Knowledge Content interpreted by a human being generates knowledge. Only human beings can have knowledge. Knowledge is created in a certain context A certain data set can generate many different information set A certain information can generate lot of different knowledge

10. © Per Flensburg 10 Semantics Does this means 5th of March 2004? Or does it means 3rd of April 2005? Or maybe even 4th of May 2003? The semantic description tells us which is the case. The semantic is the description of metadata CustomerOrdered partDay of orderQuantity StubbheadConeswinger0407072 GrimsfeldCrwth0407062 TurbinTravers0403053

11. © Per Flensburg 11 Formal semantic Customer: Person or company registered as customer in our customer file. Ordered parts: The name of the parts the customer has ordered. The names must be the same as in our part database Day of order: Year (two last digits), month (two digits) and the number of the day in the month. Number: The number of {parts} the {customer} is ordering. Sometimes this is called ontology!

12. © Per Flensburg 12 Ontology Philosophical definition: The science about reality Here we use two other definitions: A description of the reality that corresponds to certain content A formal definition of concepts in a data base All ontologies are based upon something that is known from the beginning They are also meant for sharing

13. © Per Flensburg 13 Increasing interest for ontologies

14. © Per Flensburg 14 An example CustomerOrdered partDay of orderQuantity StubbheadConeswinger0407072 GrimsfeldCrwth0407062 TurbinTravers0406063 Let us pick one line and study its ontology

15. © Per Flensburg 15 Ontology of the line CustomerOrdered partDay of orderQuantity GrimsfeldCrwth0407062

16. © Per Flensburg 16 But what does it mean? This is a description (In fact, some pictures) of the reality But how shall it be interpreted? What kind of business is it all about? One interpretation: Grimsfeld has a music shop and “we” are selling music instruments However “travers” is no music instrument. And “coneswinger” I have no idea about!

17. © Per Flensburg 17 This is a travers And Turbin ordered three of them!!

18. © Per Flensburg 18 Three travers?? It is simple: Turbin is an architect designing factories and he builds models of them. Doing so he sometime needs a model of a travers Lego has made such models Unfortunately they are not manufactored any longer But since we have an antiquity shop we sell all kinds of old things Inclusive old Lego models But still, I don’t know anything about coneswingers...

19. © Per Flensburg 19 Business system A Business system B Data Structure Data Structure Information Content Web Services XML etc, Metadata Ontology To move information and meaning WSCC CeLeKT Ontology integration

20. Ontologies from the textbook

21. © Per Flensburg 21 Sources Main source for these slides: Dieter Fensel: Ontologies: Silver Bullet for Knowledge Management and Electronic Commerce (http://www.msi.vxu.se/~per/bullet.pdf) On the home page of the course there are a lot of interesting articles about semantics and ontology. (http://www.msi.vxu.se/~per/IVC743/IVC743.html)

22. © Per Flensburg 22 Definition (från AI) In the simplest case, an ontology describes a hierarchy of concepts related by subsumption relationships; in more sophisticated cases, suitable axioms are added in order to express other relationships between concepts and to constrain their intended interpretation.

23. © Per Flensburg 23 More definitions Fensel: shared and common understanding of a domain that can be communicated between people and heterogeneous and widely spread application systems. Fensel again: ontologies describe the static domain knowledge of a knowledge-based system.

24. © Per Flensburg 24 More about ontologies A language for defining ontologies is syntactically and semantically richer than common approaches for databases. The information that is described by an ontology consists of semi-structured natural language texts and not tabular information. An ontology must be a shared and consensual terminology because it is used for information sharing and exchange. An ontology provides a domain theory and not the structure of a data container.

25. © Per Flensburg 25 Example -schedule

26. © Per Flensburg 26 Schedule - metadata Simple table structure Data base schema

27. © Per Flensburg 27 Schedule - instance A row in the schedule Is in fact a representation of a fact.

28. © Per Flensburg 28 This is a meta-meta description in relation to the fact Semantics The number of the week, according to standard ISO 321- 543-432-645.a The day expressed as weekday, number and month

29. © Per Flensburg 29 Ontology (home-made) Name: Schedule at VXU Purpose: Temporal relation between the following entities: Room, Person and Activity. Temporal expression: Week, day and time in nov- dec 02 Room-domain: All lecture rooms at Växjö university with a capacity of at least 35 persons Person-domain: PF, RL, Inge Andersson and Olle Dahlborg Activity-domain: {in fact a description of what is going to be dealt with in each lecture}

30. © Per Flensburg 30 Purpose The ontology can be used in many cases, not only for our schedule It says something about the content, not the form If you see an instance with a value not belonging to the ontology, you know something is incorrect If you are familiar with the ontology, no further explanations is needed in order to understand the meaning.

31. © Per Flensburg 31 Syntax of ontology This specific ontology has the following syntax: Name: Purpose: relation between Temporal expression: Room-domain: Person-domain: Activity-domain:

32. © Per Flensburg 32 Initiatives Resource Description Framework (RDF) Semantic web XML Schemes, standard for describing the structure and part of the semantics of data. XSL, describing mappings between different presentation sheets.

33. © Per Flensburg 33 Intranets and semantics In a fast changing world knowledge becomes increasingly important Maintaining and accessing knowledge (organisational memory) is thus important. The knowledge is often weakly structured, stored in intranets and in different formats (picture, sound etc.) Knowledge management, which turn information into useful knowledge is thus heavily needed.

34. © Per Flensburg 34 Document management in intranets Key-word based retrieval provides lots of irrelevant information out of context. Extracting information requires human attention, both for extracting and integrating Maintaining weakly structured sources is time- consuming

35. © Per Flensburg 35 Semantic possibilities Search for content, not key-words Query answering instead of information retrieval Correct exchange of structured or semi-structured information via for instance XSL. Define view on documents or sets of documents, information fusion

36. © Per Flensburg 36 Agents that find the best shopping opportunity. Example: Shop-bots B2C-site Shop-bot Wrapper

37. © Per Flensburg 37 Problems with shop-bots A wrapper is needed for each place and type of bot. No flexibility in retrieving the information Information at the B2C-site must be provided in a structured form. Usually this information is provided in natural language also which inevitably will cause inconsistency problems

38. © Per Flensburg 38 Solution Using various XML-techniques provides better possibilities for translation between the bot and the site. However, they must share the same ontology. An ontology describes the various products and can be used to navigate and search automatically for the required information.

39. © Per Flensburg 39 Electronic commerce B2B Standard techniques, such as EDIFACT cumbersome and error-prone and not integrated with other documents. The XML-family of techniques can be used for describing syntax and semantics of data, but not for the business processes and for the products. Standard ontologies in combination with XLS-based translation services is thus needed.

40. © Per Flensburg 40 Types of ontologies Domain ontologies capture the knowledge valid for a particular type of domain Metadata ontologies like Dublin Core provide a vocabulary for describing the content of on-line information sources (Libraries). Generic or common sense ontologies aim at capturing general knowledge about the world, providing basic notions and concepts for things like time, space, state, event etc.

41. © Per Flensburg 41 More types Representational ontologies provide representational entities without stating what should be represented. A well-known representational ontology is the Frame Ontology which defines concepts such as frames, slots, and slot constraints allowing the expression of knowledge in an object-oriented or frame-based way. Method and task ontologies provide a reasoning point of view on domain knowledge such as hypothesis, cause-effect statements etc.

42. © Per Flensburg 42 Constructing ontologies Prerequisite: ontologies are small modules with a high internal coherence and a limited amount of interaction between the modules. Constructing a new ontology is a matter of assembling existing ones. Inclusion Restriction Polymorphic refinement

43. © Per Flensburg 43 Formal languages Various kind of formal languages are used for representing ontologies, among others Description logics Frame Logic First-order predicate logic extended with meta- capabilities to reason about relations.

44. Some examples from Fensel

45. © Per Flensburg 45 The Sensus system The basic idea is to use so-called seed elements which represent the most important domain concepts for identifying the relevant parts of a toplevel ontology. The selected parts are then used as starting points for extending the ontology with further domain specific concepts.

46. © Per Flensburg 46 Word-Net An on-line lexical reference system. English nouns, verbs, adjectives and adverbs are organized into synonym sets, each representing one underlying lexical concept. Different semantic relations link the synonym sets. WordNet contains around 100.000 word meanings organized in a taxonomy. http://www.cogsci.princeton.edu/~wn/

47. © Per Flensburg 47 Semantical relationships Synonymy: Similarity in meaning of words. Antonymy: Dichotomy in meaning of words Hyponymy: Is-a relationship between concepts. This is-a hierarchy ensures the inheritance of properties from superconcepts to subconcepts. Meronymy: Part-of relationship between concepts. Morphological relations which are used to reduce word forms.

48. © Per Flensburg 48 Features of Word-Net Free of charge Multilingual European version also exists (http://www.let.uva.nl/~ewn ) Its large size (i.e., number of concepts) Its domain-independence Its low level of formalization The definitions are vague and limits the possibility for automatic reasoning support

49. © Per Flensburg 49 CYC http://www.cyc.com/ Comes from AI Humans decide based on their common sense knowledge what to learn and what not to learn from their observations. CYC started as an approach to formalize this knowledge and provide it with a formal and executable semantics. Hundreds of thousands of concepts have been formalized with millions of logical axioms, rules, and other assertions.

50. © Per Flensburg 50 Features of CYC The upper-level ontology of CYC with 3000 concepts has been made publicly available. Most of the more specific concepts are kept secret CYC groups concepts into microtheories to structure the overall ontology. They are a means to express context dependency i.e., what is right in one context may be wrong in another CycL, a variant of predicate logic, is used as language for expressing these theories.

51. © Per Flensburg 51 TOVE (TOronto Virtual Enterprise) Task and domain-specific ontology. The ontology supports enterprise integration, providing a shareable representation of knowledge in a generic, reusable data model TOVE provides a reusable representation (i.e., ontology) of industrial concepts. http://www.eil.utoronto.ca/tove/toveont.html

52. © Per Flensburg 52 Characteristics It provides a shared terminology for the enterprise that each agent can jointly understand and use It defines the meaning of each term in precise and unambiguous manner as possible It implements the semantics in a set of axioms that will enable TOVE to automatically deduce the answer to many “common sense” questions about the enterprise It defines a symbology for depicting a term or the concept constructed thereof in a graphical context

53. © Per Flensburg 53 (KA)2 – a case study Knowledge Annotation Initiative of Knowledge Acquisition Community http://www.aifb.uni- karlsruhe.de/WBS/broker/KA2.html The process of developing an ontology for a heterogeneous and world-wide (research) community The use of the ontology for providing semantic access to on-line information sources of this community.

54. © Per Flensburg 54 Example in (KA)2 Class: research-topic Attributes: Name: Description: Approaches: Research-groups: Researchers: Related-topics: Subtopics: Events: Journals: Projects: Application-areas: Products: Bibliographies: Mailing-lists: Webpages: International-funding-agencies: National-funding-agencies: Author-of-ontology: Date-of-last-modification:

55. © Per Flensburg 55 Procedure A lot of instances of the schema was developed and published on the home page Examples: specification languages knowledge acquisition methodologies agent-oriented approaches knowledge acquisition from natural language knowledge management

56. Knowledge management An application

57. © Per Flensburg 57 Knowledge management deals with Acquiring Maintaining Accessing knowledge of an organization. Here we will apply it to internet and concentrate on the last issue: Search for knowledge.

58. © Per Flensburg 58 Search engines They have typically three parts: A webcrawler for downloading, an indexer for finding key terms and a query interface that retrieves answers to the proposed questions. They are all based on keywords. The indexing process of the web-pages is thus crucial for the retrieval.

59. © Per Flensburg 59 Search domain Consists of about 300 millions fix documents, but this is only about 20% of what is available in total. The rest (80%) is dynamically generated (example: Aftonbladet) Altavista provides it all, Google sort according to documents pointing at the actual document and Yahoo uses human invention.

60. © Per Flensburg 60 Dimensions in searching Precision: how many retrieved documents are really relevant? Recall: have I found all relevant information? Time: for the humans to find the desired information among the retrieved. Scattering:The information might be scattered over several pages with only implicit relations between them

61. © Per Flensburg 61 Ontobroker Define an ontology Use it to annotate/structure/wrap your web documents Somebody else can make use of Ontobroker’s advanced query and inference services to consult your knowledge. To achieve this goal, Ontobroker provides three interleaved languages and two tools.

62. © Per Flensburg 62 Languages It provides a representation language formulating ontologies. A subset of it is used to formulate queries, i.e. to define the query language. An annotation language is offered to enable knowledge providers to enrich web documents with ontological information.

63. © Per Flensburg 63 Representation language

64. © Per Flensburg 64 Some definitions Class definition: c[] defines a class with name c. Attribute definition: c[a=>> {c1,...,cn}] implies that the attribute a can applied to the elements of c and an attribute value must be member of all classes c1,...,cn. Is-a relationship: c1:: c2 defines c1 as a subclass of c2 which implies that: all elements of c1 are also elements of c2 all attributes and their value restrictions defined for c2 are also defined for c1, and multiple attribute inheritance exists

65. © Per Flensburg 65 More definitions Is-element-of relationship: e : c defines e as an element of the class c. Rules like FORALL x,y x[a ->> y] > x]. If a is an attribute for x it is also an attribute for y FORALL x,y x:c1[a1 ->> y] y:c2[a2 ->> x]. The common set of attributes for x and y

66. © Per Flensburg 66 Annotation language Ontobroker provides an annotation language called HTMLA The following HTML page states that the text string „Richard Benjamins“ is the name of a researcher where the URL of his homepage is used as his object id. Welcome to my homepge My name is Richard Benjamins. Cf XML!

67. © Per Flensburg 67 Query language The query language is defined as a subset of the representation language. The elementary expression is: X ∈ c Λ attribute(x) = v written in Frame logic: x[attribute -> v] : c

68. © Per Flensburg 68 Inference engine

69. © Per Flensburg 69 Summary Ontobroker only recognises pages that are annotated according to its rules. Principally you still have a database and a database schema in the bottom What can’t be expressed in first order predicat logic can’t be expressed.

70. © Per Flensburg 70 New idea (On2broker) XML, RDF etc. The webOntologies Information agent Fact retrieval Data baseQuery engine

71. © Per Flensburg 71 Cf, Skogsresurs Concept base Search agent URL Classification URL-database Profile Interesting links User

72. © Per Flensburg 72 Comments Facts are defined by the logic of the ontologies They are retrieved according to logical rules They are stored into formal databases Thus only facts possible to express in first order predicat logic is possible to retrieve. Also On2broker has severe efficiency problems if the number of facts extend 100 000

73. © Per Flensburg 73 IBROW: Dynamic reasoning Brokering dynamic reasoning services in WWW http://www.swi.psy.uva.nl/projects/IBROW3/home. html It will access libraries in the Internet, search for appropriate inference services, verify their requirements, request additional information from the customer if needed, adapt the inference services to the particular domain knowledge, plug them together, and execute them via CORBA. Therefore, the user no longer buys, downloads and installs software. Instead he uses it as a computational service provided via the network

74. © Per Flensburg 74 Use of IBROW In a business-to-business (B2B) context, IBROW technology can be used to construct half products, which then need further processing by industries before delivering end products to consumers. For example, a car manufacturer could be interested in a service that helps him to develop and/or adapt a new car design. In another scenario, the IBROW broker provides a service to configure the bare bones of a knowledge system, which then needs to be refined for end consumers based on their particular needs.

75. © Per Flensburg 75 Other use Yet another model would use IBROW technology to provide an underlying infrastructure to support knowledge engineers in selecting, testing, adapting, refining, and combining generic components into concrete systems.

76. Electronic Commerce An application are for ontologies

77. © Per Flensburg 77 Types of products Intelligent information search agents (i.e., shopping agents) that help customers to find products. Intelligent information providers (i.e., on-line stores) that help vendors to present their goods in appropriate manner. Intelligent information brokers (i.e., on-line market places) that mediate between buyers and vendors.

78. © Per Flensburg 78 Shopbots Client (Browser) Shopbot Wrapper 1 On-line store 1 Wrapper 2 On-line store 2 Wrapper 3 On-line store 3

79. © Per Flensburg 79 Examples Bookblvd (http://www.bookblvd.com/ ), Bottom Dollar (http://www.bottomdollar.com/ ), Buyer’s Index (http://www.buyersindex.com/ ), CompareNet (http://www.compare.net/ ), Dealpilot (http://www.dealpilot.com/ ), Jango (http://www.jango.com/ ), Junglee (http://www.junglee.com/) MyShop (http://www.myshop.de ), Shopfind (http://www.shopfind.com/ ) Shopper (http://www.shopper.com ).

80. © Per Flensburg 80 Why shopbots fail Web users do not want to pay because they are used to free service. Product providers do not want to fund the agent because of its ability to always find the cheapest source. Product providers would fund the agent if it manipulated the search results. This would eliminate objectivity however which is a requirement for high acceptance. In the end, most of them were bought by Internet portals

81. © Per Flensburg 81 Other types of bots Lifestyle finder recommends documents matching your interests based on your answers to a set of questions. Alexa: watches all its users and anticipates their preferences. An “intelligent” proxy which caches the pages a user will visit. Firefly asks for ratings of specific musical artists, correlates each user with others who share their tastes, and recommends songs or albums which their cohorts have rated highly.

82. © Per Flensburg 82 1:1, negotiation, often EDIFACT-based 1:N, A large company dictates, often EDIFACT N:M, a fragmented marketplace B2B

83. © Per Flensburg 83 The fragmented marketplace Replace a trade agent Bring seller and buyer together on global base Require means for translation and representation Require means for content description

84. © Per Flensburg 84 Translation

85. © Per Flensburg 85 Representation and translation RDF provides a standard for describing semantics XML provides a standard for describing the structure of a document XML schema provides a standard for describing the semantics XSL provides a standard for describing mappings between terminologies But none provides a standard vocabulary

86. © Per Flensburg 86 Ontologies in B2B Corresponds to standardised product catalogues. Attempts: Common business library (http://www.commerce.net) Commerce XML (http://www.oasis-open.org ) Dublin Core (http://www.indecs.org ) Rosettanet (http://www.rosettanet.org/ ) Etc.

87. © Per Flensburg 87 Branch portals Chemdex (www.chemdex.com) Life science products trading PaperExchange (www.paperexchange.com) Paper industry spot market VerticalNet (www.verticalnet.com) Generic portal

88. © Per Flensburg 88 VerticalNet Seller ontology 1 Seller ontology 2 Seller ontology 3 Buyer ontology 1 Buyer ontology 2 Buyer ontology 3 Transformation Visualisation Standard ontology Ontology mapping

89. Fine Thanx...

90. To sum it all up Putting it all together

91. © Per Flensburg 91 The techniques XML: Describing the structure of a document, cf. defining layout in a database DTD: Element declaration that define composed tags and value ranges for elementary tags Attribute declaration that define attributes of tags Entity declaration XSL: Describes how to render the document and the processing of the parts there, cf. calculating fields in a database. Can be used for translation

92. © Per Flensburg 92 The role of XSL XML DTD 1 XML DTD 2 XSL

93. © Per Flensburg 93 A document XML DTD XSL Name:Value on name Part:Value on part Quantity:Value on quant Price:Value on price Part:Value on part Quantity:Value on quant Price: Value on price Total price: Value on total Adress:Value on adr Ontololgy

94. © Per Flensburg 94 XML and semantics XML describe documents in tree format Thus the overlaying context is transferred to the underlying branches This is the only semantics that exists.

95. © Per Flensburg 95 RDF A subject is an entity that can be referred to by a address in the WWW (i.e., by an URI). A predicate defines a binary relation between resources and/or atomic values provided by primitive data type definitions in XML. An object specifies for a subject a value for a predicate. That is, objects provide the actual characterizations of the Web documents.

96. © Per Flensburg 96 Example of RDF Author(http://www.msi.vxu.se/~per/) = X Name(X) = Per Flensburg Email(X) = per.flensburg@msi.vxu.se Claim(Anna)=(affiliation(Author( (http://www.msi.vxu.se/~per/))=Växjö university)

97. Fine