Download Intelligent Behaviour - People

Document related concepts

Neuroinformatics wikipedia , lookup

Expert system wikipedia , lookup

Upper ontology wikipedia , lookup

Personal knowledge base wikipedia , lookup

Transcript
Exploiting large scale web semantics
to build end user applications
Enrico Motta
Professor of Knowledge Technologies
Knowledge Media Institute
The Open University
Aims of the Talk
• What is the Semantic Web
– Perspectives
• The SW as a ‘web of data’
• The SW as a new context in which to build semantic
applications and an unprecedented opportunity in
which to address some classic AI problems
– Typical misconceptions
• What the SW is not!
• Semantic Web for Users
– Applications that do something interesting and useful
to users, by exploiting available web semantics
The Semantic Web as a ‘Web of Data’
Making data available to SW-aware software
<foaf:Person rdf:about="http://identifiers.kmi.open.ac.uk/people/enrico-motta/">
<foaf:name>Enrico Motta</foaf:name>
<foaf:firstName>Enrico</foaf:firstName>
<foaf:surname>Motta</foaf:surname>
<foaf:phone rdf:resource="tel:+44-(0)1908-653506"/>
<foaf:homepage rdf:resource="http://kmi.open.ac.uk/people/motta/"/>
<foaf:workplaceHomepage rdf:resource="http://kmi.open.ac.uk/"/>
<foaf:depiction rdf:resource="http://kmi.open.ac.uk/img/members/enrico.jpg"/>
<foaf:topic_interest>Knowledge Technologies</foaf:topic_interest>
<foaf:topic_interest>Semantic Web</foaf:topic_interest>
<foaf:topic_interest>Ontologies</foaf:topic_interest>
<foaf:topic_interest>Problem Solving Methods</foaf:topic_interest>
<foaf:topic_interest>Knowledge Modelling</foaf:topic_interest>
<foaf:topic_interest>Knowledge Management</foaf:topic_interest>
<foaf:based_near>
<geo:Point>
<geo:lat>52.024868</geo:lat>
<geo:long>-0.707143</geo:long>
<contact:nearestAirport>
<airport:name>London Luton Airport</airport:name>
<airport:iataCode>LTN</airport:iataCode>
<airport:location>Luton, United Kingdom</airport:location>
<geo:lat>51.866666666667</geo:lat>
<geo:long>-0.36666666666667</geo:long>
<rdfs:seeAlso rdf:resource="http://www.daml.org/cgi-bin/airport?LTN"/>
<foaf:currentProject>
<foaf:Project>
<foaf:name>AquaLog</foaf:name>
The web of SW documents
Current status of the
semantic web
•
10-20 million semantic web documents
– Expressed in RDF, OWL, DAML+OIL
•
7K-10K ontologies
– These cover a variety of domains - music,
multimedia, computing, management, bio-medical
sciences, upper level concepts, etc…
•
Hence:
– To a significant extent the semantic web is already
in place
– However, domain coverage is very uneven
– Still primarily a research enterprise, however
interest is rapidly increasing in both governmental
and business organizations
• “early adopters” phase
The above figures refer to resources which are publicly
accessible on the web
Bibliographic Data
CS Dept Data
Geography
AKT Reference Ontology
<rdf :Description
rdf :abo ut=" ht t p:/ /ww w.ecs. sot on.ac.uk/info/#p erso
n-01 2 6 9 ">
<ns 0 :family -name>Gibbins</ns0:family -name>
<ns 0 :full -name>Nicholas Gibbins</ns0:full
-name>
<ns 0 :given-name>N icholas</ns0:g iven-name>
<ns 0 :has-email address> [email protected] on.ac.u k</ns0:has -email address>
<ns 0 :has-affi liation -to -unit
rdf: resour ce=" ht t p:// 1 94 .66 .1 8 3. 2 6 / WEBSITE/G OW
/Vie wDepartme nt.aspx?Dep art ment =7 5 0"/>
< / rdf :Descriptio n>
</ rdf :RDF>
RDF Data
“Corporate Semantic Webs”
• A ‘corporate ontology’ is used to provide a
homogeneous view over heterogeneous data
sources.
• Often tackle Enterprise Information Integration
scenarios
• Hailed by Gartner as one of the key emerging
strategic technology trends
– E.g., Garlik is a multi-million startup recently set up in UK to support
personal information management, which uses an ontology to
integrate data mined from the web on a large scale
AquaLog
Applications that exploit
large scale semantic content
The web of data
Gateways to the SW
Application
Semantic
Web
• Sophisticated quality control mechanism
– Detects duplications
– Fixes obvious syntax problems
• E.g., duplicated ontology IDs, namespaces, etc..
• Structures ontologies in a network
– Using relations such as: extends, inconsistentWith, duplicates
• Provides interfaces for both human users and software
programs
• Provides efficient API
• Supports formal queries (SPARQL)
• Variety of ontology ranking mechanisms
• Modularization/Combination support
• Plug-ins for Protégé and NeOn Toolkit
• Very cool logo!
Case Study 1: Automatic Alignment of Thesauri
in the Agricultural/Fishery Domain
Method
- SCARLET - matching
by Harvesting the SW
- Automatically select
and combine multiple
online ontologies to

derive a relation
Access
Semantic Web

Scarlet
Deduce
Concept_A
(e.g., Supermarket)
Semantic Relation
Concept_B
)
(e.g., Building)
(
Two strategies

Building
OrganicChemical

PublicBuilding




Shop

Supermarket
Steroid
Lipid
Steroid
Cholesterol
Semantic Web




Scarlet
Supermarket

Building
Scarlet
Cholesterol

OrganicChemical
(A)
(B)
Deriving relations from (A) one ontology and (B) across ontologies.
Experiment
Matching:
• AGROVOC
•UN’s Food and
Agriculture
Organisation (FAO)
thesaurus
•28.174 descriptor terms
•10.028 non-descriptor
terms
• NALT
•US National Agricultural
Library Thesaurus
•41.577 descriptor terms
•24.525 non-descriptor
terms
226 Used Ontologies
http://139.91.183.30:9090/RDF/VRP/Examples/tap.rdf
http://reliant.teknowledge.com/DAML/SUMO.daml
http://reliant.teknowledge.com/DAML/Mid-level-ontology.daml
http://gate.ac.uk/projects/
htechsight/Technologies.daml
http://reliant.teknowledge.com/DAML/Economy.daml
Evaluation 1 - Precision
• Manual assessment of 1000 mappings (15%)
• Evaluators:
– Researchers in the area of the Semantic Web
– 6 people split in two groups
• Results:
– Comparable to best results for background
knowledge based matchers.
Evaluation 2 – Error Analysis
Case Study 2:
Folksonomy Tagspace Enrichment
Features of Web2.0 sites
• Tagging as opposed to rigid
classification
• Dynamic vocabulary does
not require much annotation
effort and evolves easily
• Shared vocabulary emerge
over time
– certain tags become
particularly popular
Limitations of tagging
• Different granularity of tagging
– rome vs colosseum vs roman monument
– Flower vs tulip
– Etc..
• Multilinguality
• Spelling errors, different terminology, plural vs
singular, etc…
• This has a number of negative implications for
the effective use of tagged resources
– e.g., Search exhibits very poor recall
Giving meaning to tags
What does it mean to add
semantics to tags?
1. Mapping a tag to a SW
element
"japan"
<akt:Country Japan>
2. Linking two "SW tags" using semantic relations
{japan, asia}
<japan subRegionOf asia>
Applications of the approach
• To improve recall in keyword search
• To support annotation by dynamically
suggesting relevant tags or visualizing the
structure of relevant tags
• To enable formal queries over a space of tags
– Hence, going beyond keyword search
• To support new forms of intelligent navigation
– i.e., using the 'semantic layer' to support navigation
Pre-processing
Folksonomy
Tags
Clustering
Clean tags
Analyze co-occurrence
of tags
Group similar tags
Co-occurence matrix
Filter infrequent tags
Cluster tags
Concise tags
Cluster1
Yes
Cluster2
…
Clustern
2 “related” tags
SW search engine
Remaining
tags?
No
Find mappings &
relation for pair of tags
Wikipedia
Google
END
<concept, relation,
concept>
Concept and relation identification
Examples
Cluster_1: {admin application archive collection component
control developer dom example form innovation interface
layout planning program repository resource sourcecode}
Information Object
archive
has-mention-of
participant
creator
developer
participatesIn
in-event
event
activity
innovation
application
planning
resource
typeRange
example
user
component
admin interface
Examples
Cluster_2: {college commerce corporate course education high
instructing learn learning lms school student}
activities4
learning4
teaching4
education
training1,4
qualification
school2
corporate1
institution
postSecondary
School2
student3
studiesAt
takesCourse
university2,3
offersCourse
course3
1http://gate.ac.uk/projects/htechsight/Employment.daml.
2http://reliant.teknowledge.com/DAML/Mid-level-ontology.daml.
3http://www.mondeca.com/owl/moses/ita.owl.
4http://www.cs.utexas.edu/users/mfkb/RKF/tree/CLib-core-office.owl.
college2
Faceted Ontology
• Ontology creation and
maintenance is
automated
• Ontology evolution is
driven by task features
and by user changes
• Large scale integration of
ontology elements from
massively distributed
online ontologies
• Very different from
traditional top-downdesigned ontologies
Case Study 3:
Reviewing and Rating on the Web
Revyu.com
Trust Factors
expertise
the source has relevant expertise of the domain of the
recommendation-seeking; this may be formally validated through
qualifications or acquired over time.
experience
the source has experience of solving similar scenarios in this domain,
but without extensive expertise.
impartiality
the source does not have vested interests in a particular resolution to
the scenario.
affinity
the source has characteristics in common with the recommendation
seeker, such as shared tastes, standards, values, viewpoints,
interests, or expectations.
track record
the source has previously provided successful recommendations to
the recommendation seeker.
solution
subjective
objective
affinity
expertise
experience
factors
emphasised
Applying the framework to
revyu.com
• Affinity
– Operationalised as the degree of overlap in items
reviewed, and in ratings given
• Experience
– Proxy metric: Usage of particular tags (as proxies for
topics)
• Experience scores based on tagging data
• Integrates also data from del.icio.us for those users
who have chosen to publish their del.icio.us account on
FOAF
• Expertise
– Proxy metric: Credibility
– Captures the social aspect of expertise: endorsement
Using trust factors for
ranking reviews
PowerAqua and PowerMagpie
How does the Semantic Web relate to
Artificial Intelligence research?
AI as Heuristic Search
The knowledge-based
paradigm in AI
“Today there has been a shift in paradigm. The
fundamental problem of understanding
intelligence is not the identification of a few
powerful techniques, but rather the question of
how to represent large amounts of knowledge in
a fashion that permits their effective use”
Goldstein and Papert,1977
Knowledge Representation
Hypothesis in AI
Any mechanically embodied intelligent process
will be comprised of structural ingredients that
we as external observers naturally take to
represent a propositional account of the
knowledge that the overall process exhibits, and
independent of such external semantic
attribution, play a formal but causal and
essential role in engendering the behaviour that
manifests that knowledge
Brian Smith, 1982
Knowledge-Based Systems
Large Body
of Knowledge
Intelligent Behaviour
The Knowledge Acquisition
Bottleneck
Knowledge
Large Body
of Knowledge
KA
Bottleneck
Intelligent Behaviour
The Cyc project
Structured libraries of
reusable components
Problem Solving
Method
Generic Task
Library of PSMs
Parametric Design
Classification
Scheduling
Etc…
Mapping
Knowledge
Mapping Ontology
Application-specific
Problem-Solving Knowledge
Ontology
Application Configuration
Domain Model
The next knowledge medium
• However, our approach based on structured
libraries of problem solving components only
addressed the economic cost of KBS
development…
“An information network with semi-automated
services for the generation, distribution, and
consumption of knowledge”
SW as Enabler of Intelligent
Behaviour
Both a platform for
knowledge publishing
and a large scale source
of knowledge
Intelligent Behaviour
KBS vs SW Systems
Classic KBS
SW Systems
Provenance
Centralized
Distributed
Size
Small/Medium
Extra Huge
Repr. Schema
Homogeneous
Heterogeneous
Quality
High
Very Variable
Degree of trust
High
Very Variable
Key Paradigm Shift
Intelligence
Classic KBS
SW Systems
A function of
sophisticated,
logical, taskcentric problem
solving
A side-effect of
being able to
integrate
different types of
reasoning to
handle size and
heterogeneous
quality and
representation
Conclusions
Typical misconceptions…
• “The SW is a long-term vision…”
– Ehm…actually… it already exists…
• “The SW will never work because nobody is going to
annotate their web pages”
– The SW is not about annotating web pages, the SW is
a web of data, most of which are generated from DBs, or
from web mining software, or from applications which
produce SW technology
• “The idea of a universal ontology has failed before and
will fail again. Hence the SW is doomed”
– The SW is not about a single universal ontology.
Already there are around 10K ontologies and the number
is growing…
– SW applications may use 1, 2, 3, or even hundreds of
ontologies.
Large Scale Distributed
Semantics
• Widespread production of formalised knowledge models
(ontologies and metadata), from a variety of different
groups and individuals
– E.g., legal, bio-medical, governmental, environmental,
music, art, multimedia, computing, etc..
– “Knowledge modelling to become a new form of literacy?”
• Stutt and Motta, 1997
• This large scale heterogenous resource will enable a
new generation of semantic-aware technologies
• These developments may provide a new context in
which to address the economic barriers to KBS
development
• The SW already exists to some extent, however there is
still a way to go, before it will reach the required degree
of maturity
Large Scale Distributed
Semantics
• Much like AI, the semantic web will only
succeed if it becomes ubiquitous and hidden
“There's this stupid myth out there that A.I. has failed, but A.I. is
everywhere around you every second of the day. People just don't
notice it. You've got A.I. systems in cars, tuning the parameters of
the fuel injection systems. When you land in an airplane, your
gate gets chosen by an A.I. scheduling system. Every time you
use a piece of Microsoft software, you've got an A.I. system trying
to figure out what you're doing, like writing a letter, and it does a
pretty damned good job. Every time you see a movie with
computer-generated characters, they're all little A.I. characters
behaving as a group. Every time you play a video game, you're
playing against an A.I. system.”
Rodney Brooks