Transcript Practical Knowledge Representation for the Web

PRACTICAL KNOWLEDGE
REPRESENTATION FOR
THE WEB
Frank van Harmelen Dieter Fensel AIFB
Kim Kangil
Structural Complexity Laboratory
CONTENTS

Existing Semantic mark-up languages

Symbol level comparison

Knowledge level comparison

conclusion
EXISTING SEMANTIC MARKUP
LANGUAGES

HTML based
HTML <META> - tags
 SHOE


HTML derived
HTML <SPAN> - elements ; Ontobroker
 Cascading Style Sheets

XML
 RDF

HTML <META>-TAGS
SHOE
HTML <SPAN>-ELEMENTS
ONTOBROKER
CASCADING STYLE SHEETS
PROPERTIES

<meta>-tag
Global property
 Anchor mechanism


SHOE
Extension of meta tag
 Independent to location
 Arbitrary relation number
 Global property


<span>-elements
Standard CLASS attribute,
 Structure for giving semantic is in a document


Ontobroker
First order logic
 Similar to span-tag


CSS
to separate structure information
 Style specification
 Can be used for adding semantic information

XML
-A Labeled tree
-Nesting
-DTD
RDF
-Data model
-Resource
-Property
-Statement
-No assumption to add structural
information ?
- XML schema base
SYMBOL COMPARISON

Support of web technology

Avoiding duplication

Allowing nesting
AVOIDING DUPLICATION
Removing the duplicated expression to add
semantic information and to render that
 Reducing the cost to use in the Web

<meta> : non-standard, anchor mechanism
 <SHOE> : duplicated
 <XML>,<SPAN-tag> : use same information for
rendering and adding semantic information
 <RDF> : separation is intened

ALLOWING NESTING &
SUPPORTING WEB TECHNOLOGY

In language design, Nesting of expression is the
typcial way to use the scoping
Xml, CSS, <SPAN> can support nesting
 RDF can’t support nesting with natural way


Supporting web technology means how well
some languages can be spread out for using AI on
the web

Syntactical variety can be harmful to be supported,
ex. SHOE,Ontobroker
KNOWLEDGE COMPARISON

Factual knowledge : Data–model

Terminological knowledge : ontology

Inferential knowledge
FACTUAL KNOWLEDGE : DATA
MODEL

Various data model type





Meta-tag : basic html attribute mechanism
XML & Span-tag : labeled trees
Ontobroker : expression in F-logic , complicated expression
could be included to onto- attribute. Multiple inheritance of
attributes.
RDF : use binary relation. extended by reificatoin
SHOE : n-ary relation. Use attribute for classes, multiple
inheritance of attributes
RDF, Ontobroker, SHOE support object oriented type
schema
 RDF is property-centric ( don’t use attribute ) – don’t refine
when it is inherited to sub classes. Sharing property is
impossible

TERMINOLOGICAL KNOWLEDGE :
ONTOLOGY





Specification of conceptualization vocabulary to
describe domain.
Meta,span-tag is not making data-schema separately.
CSS make it , but just a list of category names.
SHOE, Ontobroker provide explicit ontology.
Ontobroker has single centrally defined ontology. But
SHOE could extend ontology locally.
DTD of xml is close to ontology,but just lexical nesting
specification.


Missing : ontological hierachical specification, inheritance
mechanism, range restrictions on attribute.
RDF can describe ontology but, it needs reification
INFERENCE KNOWLEDGE
SHOE : pure HORN rules
 Ontobroker : first order logic fragments
 Other things impossible. ?

CONCLUSION
On the symbol comparison , span-tag has much of
functionality of XML
 On the knowledge comparison, META and SPANtag is not rich. Surprising thing is, RDF and XML
also don’t support to use ontology and inference
 For using AI on realistic, large-scale Web
application, Span-tag will good to support it.
 RDF needs more development for representing
ontology, inferential knowledge
 Essentially, SHOE, Ontobroker ,these two AI
based language is useful on the knowledge level
feature
