Transcript Talk presented at Knowledge Representation 2002

KR: Reinjecting Reality
Mathematical ideas originate in empirics .. But, once they
are so conceived, the subject begins to live a peculiar life of
its own and is better compared to a creative one, governed
almost entirely by aesthetical motivations …As a
mathematical discipline travels, or after much abstract
inbreeding, [it] is in danger of degeneration…whenever this
stage is reached, the only remedy seems to me to be the
rejuvenating return to the source; the reinjection of more or
less directly empirical ideas
--- John Von Neumann, 1953
KR2002, Apr 2002
1
The Semantic Web:
KR’s Worst Nightmare?
Professor James Hendler
http://www.cs.umd.edu/~hendler
Co-Director, Maryland Information and Network
Dynamics Laboratory
The nightmare: KR becomes relevant
Artificial Intelligence researchers have studied such
systems since long before the web was developed.
Knowledge representation, as this technology is often
called, is currently in a state comparable to that of
hypertext before the advent of the web: it is clearly a good
idea, and some very nice demonstrations exist, but it has
not yet changed the world. It contains the seeds of
important applications, but to unleash its full power it must
be linked into a single global system.
-- Tim Berners-Lee, inventor of the WWW, 2001.
KR2002, Apr 2002
3
Outline
 The SEMANTIC web
 The semantic WEB
 We’ve heard this kind of crap before, why
should we believe this one?
 Challenges ahead
 But is it AI?
KR2002, Apr 2002
4
The SEMANTIC Web
Event:
title
Event:
date
Event:
Loc
< > a photo:Photograph,
Photo:File http://…/images#image1,
Photo:topic :event1#event:loc.
Event1 a Event:event;
Event:date “April 22-25,2002”,
Event:Loc http://…/Toulouse,
Event:Title “Eighth…”.
KR2002, Apr 2002
5
KR on the Web
 Many characteristics of the Web violate traditional
KR assumptions!
It's Large and It Grows Fast
 High Variety in Quality of Knowledge
 Diversity of Content
 Unknown/unpredictable Use Scenarios for the Knowledge
 Problems of Trust, No Single Authority
 Lack of Referential Integrity
 Knowledge acquired, not engineered

(Van Harmelen, 2000)
KR2002, Apr 2002
6
Web Semantics
Semantic Web LayerCake (Berners-Lee, 99;Swartz-Hendler, 2001)
KR2002, Apr 2002
7
Putting semantics on the web
KR2002, Apr 2002
8
(and making it machine-readable)
KR2002, Apr 2002
9
Can’t we just use XML?
This is what a web-page in natural language
looks like for a machine
XML helps
XML allows “meaningful tags” to be added to
parts of the text
< name >
< education>
< CV >
< work>
< private >
XML  machine accessible meaning
But to your machine,
the tags look like this….
name >
< name
<education>
< education>
< CV
CV >
<work>
< work>
<private>
< private >
Schemas take a step in the right direction
Schemas help….
< CV >
private
…by relating
common terms
between documents
But other people use other schemas
Someone else has one like this….
<name>
name >
<<educ>
education>
<> >
< work
<<>
private >
CV> >
<<CV
The “semantics” isn’t there
< CV >
private
…which don’t fit in
KR provides “external” referents to merge on
nme
CV
CV
work
vate
educ
ed
uc
SW languages add mappings
And structure.
CV
Which is what the web was meant to be!!

"This is a pity, as in fact documents on the web describe real objects and imaginary concepts, and
give particular relationships between them... For example, a document might describe a person.
The title document to a house describes a house and also the ownership relation with a person. ...
This means that machines, as well as people operating on the web of information, can do real
things. For example, a program could search for a house and negotiate transfer of ownership of
the house to a new owner. The land registry guarantees that the title actually represents reality.”
 Tim Berners-Lee plenary presentation at WWW Geneva, 1994
KR2002, Apr 2002
17
The semantic WEB
Goal: do to ontologies what the web does for documents
Functional
genomics Tissue
Structural
Genomics
Disease
Population
Genetics
Genome Clinical Data
Clinical trial
sequence
(Genome World - from Goble, 01)
KR2002, Apr 2002
18
This leads to a radically new view of interoperation
= some partial mapping
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
uses
Distributed,partially
mapped,
inconsistent -- but very flexible!
uses
uses
KR2002, Apr 2002
19
But, like the web…
KR2002, Apr 2002
20
Real examples
 Examples from
http://dormouse.cs.umd.edu:8080/wiki/cmsc498wiki.wiki
 Students violated every rule in the KR book



Extended existing ontologies
Linked instances directly to terms from multiple ontologies
Mixed “real KR” and NL
 We can learn from their lessons

http://dormouse.cs.umd.edu:8080/wiki/assignment1_collected_les.wiki
KR2002, Apr 2002
21
But will it fly
 DAML+OIL is probably the most used AI language ever!!

http://www.daml.org
 Gaining acceptance by web players


Semantic Web Track being offered at WWW 2002
More people will attend WWW2002 Developer Day on SW than attend KR
 Significant (international) Govt Support

US DARPA/NSF; EU IST Framework 5,6

Japan, Germany, Australia considering significant investments
US National Cancer Institute to publish cancer vocabulary in DAML+OIL

 Much New Startup activity (even in this economic climate)
 Many tools being developed

Many of them aimed at developers, not just AI literate types
KR2002, Apr 2002
22
W3C Web Ont WG
 Current Working Group includes over 50 members from 30+ organizations.

Industry including:
 Large companies such as Sun, IBM, HP, Intel, EDS, Fujitsu, Lucent, Nokia, Philips
Electronics, Unisys, Daimler0Chrysler
 Newer/smaller companies such as IVIS Group, Network Inference, Stilo Technology,
Unicorn Solutions

Government and Not-For-Profits:
 US Defense Information Systems Agency, Interoperability Technology Association for
Information Processing, Japan (INTAP) , Electricite De France, Mitre

Universities and Research Centers:
 University of Bristol, University of Maryland, University of Southamptom, Stanford
University
 DFKI (German Research Center for Artificial Intelligence), Forschungszentrum Informatik,
Ontoweb

Invited Experts (From non-W3C members)
 Well-known KR researchers (Hayes, Stein)
 Tool Developers (Dean, Heflin)
 Domain experts (Borden)

W3C Team
 Connolly (HTML, XML. XML-schema); Brickley (RDF, RDF Core)
KR2002, Apr 2002
23
Moving to the futureof the web
Semantic Web LayerCake (Berners-Lee, 99;Swartz-Hendler, 2001)
KR2002, Apr 2002
24
Web “travel agents”
How many cows are there in Texas?
Query processed: 73 answers found

Google document search finds 235,312 possible page hits.

Http://www…/CowTexas.html claims the answer is 289,921,836

A database entitled “Texas Cattle Association” can be queried for the
answer, but you will need “authorization as a state employee.”

A computer program that can compute that number is offered by the State of
Texas Cattleman’s Cooperative, click here to run program.
...
The “sex network” can answer anything that troubles you, click here for
relief...



The “UFO network” claims the “all cows in Texas have been replaced by
aliens
KR2002, Apr 2002
25
Web Agents need Service
Descriptions
KR2002, Apr 2002
26
Services need Web Logics
KR2002, Apr 2002
27
Web of Trust
 Claims can be verified if there is supporting evidence from
another (trusted) source
 We only believe that someone is a professor at a university if
the university also claims that person is a professor, and the
university is on a list I trust.
believe(c1) :- claims(x, c1) ^ predicate(c1, professorAt) ^
arg1(c1, x) ^ arg2(c1, y) ^ claims(c2, y) ^
predicate(c2, professorAt) ^ arg1(c2, x) ^
arg2(c2, y) ^ AccreditedUniversity(y)
AcknowledgedUniversity(u) :- link-from(“http://www.cs.umd.edu/university-list”,u)
Notice this one
KR2002, Apr 2002
28
Validation sites
Buy into your favorite rule set
believable(x) :- claims(src,x) ^
accreditedbyChristianCoalition(src)
 believable(x) :- claims(src,x) ^
linkfromMomsPage(src)
 believable(x) :- claims(src,x) ^
accreditedby(“http://foo.com/Unabomber/Friends/rules”,src) ^

Not-accreditedbyChristianColation(x)
KR2002, Apr 2002
29
But is it AI ?
 What about human intelligence







It's Large and It Grows Fast
Lack of Referential Integrity
High Variety in Quality of Knowledge
Diversity of Content
Unknown/unpredictable Use Scenarios for the Knowledge
Problems of Trust, No Single Authority
Knowledge acquired, not engineered
 Many characteristics of human intelligence violate
traditional KR assumptions

It’s time for us to face up to the real challenge!!
KR2002, Apr 2002
30
Conclusion
 It is no longer a question of whether the semantic web could
come into being, it can and will
 We’re already well past the starting gate



Web ontologies, term languages, “shims” to DB and services, research in
proofs/rules/trust
Standardization providing a common denominator for KR researchers as
well as web developers
Small companies starting to form, Big companies starting to move
 The KR community has lots to offer
 If, and maybe only if, it is willing to revisit some basic assumptions
 The current environment is open, encouraging, moving fast,
and exciting as heck

Come play!
KR2002, Apr 2002
31