1 building intelligent systems marie desjardins tim finin anupam joshi yun peng yelena yesha...
DESCRIPTION
UMBC an Honors University in Maryland 3 ebiquity Group Intelligent Information Systems Networking & Systems Security AIDB semantic web mobility pervasive computing trust privacy assurance web services user modeling wireless data management service discovery and composition multiagent systems machine learning policy KRTRANSCRIPT
11
Building Building Intelligent Intelligent SystemsSystems
Marie desJardinsTim Finin
Anupam JoshiYun Peng
Yelena Yesha
September 2004
tell
register
tell
register
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 22
OverviewOverview Faculty: Finin, Yesha, Joshi, Peng, desJardinsFaculty: Finin, Yesha, Joshi, Peng, desJardins Students: ~12 PhD, ~12 MS, ~5 undergradStudents: ~12 PhD, ~12 MS, ~5 undergrad Labs: Labs:
Ebiquity:Ebiquity: agents and the semantic web for agents and the semantic web for open, heterogeneous, distributed systemsopen, heterogeneous, distributed systems
MAPLE:MAPLE: MultiAgent Planning and LEarning MultiAgent Planning and LEarning FundingFunding
Current: DARPA (DAML), NSF (three ITRs, …), Current: DARPA (DAML), NSF (three ITRs, …), Intelligence community, NASA, NIST, Industry, …Intelligence community, NASA, NIST, Industry, …
Recent: DARPA (CoABS, GENOA II)Recent: DARPA (CoABS, GENOA II)
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 33
ebiquity Groupebiquity Group
IntelligentIntelligentInformationInformation
SystemsSystems
NetworkingNetworking& Systems& Systems SecuritySecurity
AIAI DBDB
semanticsemanticwebweb
mobilitymobilitypervasivepervasivecomputingcomputing trusttrustprivacyprivacy
assuranceassurance
web servicesweb services
userusermodelinmodelin
gg
wirelesswireless
data management
service service discovery and discovery and compositioncomposition
multiagentmultiagentsystemssystems
machine learningmachine learning
policypolicy
KRKR
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 44
ebiquity Groupebiquity Group
IntelligentInformation
Systems
Networking& Systems Security
AI DB
semanticweb
mobilitypervasivecomputing trustprivacy
assurance
web services
usermodeling
wireless
data management
service discovery and composition
multiagentsystems
machine learning
policy
KR
Building intelligent Building intelligent systems in open, systems in open, heterogeneous,heterogeneous,
dynamic,dynamic,distributeddistributed
environmentsenvironments
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 55
Some Relevant Project AreasSome Relevant Project Areas
(1)(1) Agents and the semantic web Agents and the semantic web(2)(2) Policies for controlling autonomous Policies for controlling autonomous
agentsagents(3)(3) Context aware pervasive computing Context aware pervasive computing(4)(4) TIVO for mobile computing TIVO for mobile computing(5)(5) Trust in information systems Trust in information systems(6)(6) Large scale semantic web systems Large scale semantic web systems
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 66
(1)(1) The Celebrity The Celebrity CoupleCouple
SemanticSemanticWebWeb
SoftwareSoftwareAgentsAgents
In 2002, Geek Gossip gushed “The semantic web will provide content for internet agents, and agents will make the semantic web “come alive”. Looks like a match made in Heaven!”
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 77
(1)(1) Trading Agents Trading Agents We’ve built an agent-based environment inspired by We’ve built an agent-based environment inspired by
TAC, the Trading Agent CompetitionTAC, the Trading Agent Competition TAC is a forum for dynamic trading agent research TAC is a forum for dynamic trading agent research
with games run in the last five yearswith games run in the last five years TAC Classic involves a travel procurement, with agents TAC Classic involves a travel procurement, with agents
buying and selling goods for clients and scored on the buying and selling goods for clients and scored on the cost and clients’ preferences for trips assembled. cost and clients’ preferences for trips assembled.
TAC is organized around a central auction serverTAC is organized around a central auction server Our goal was to open up the system, allowing peer-Our goal was to open up the system, allowing peer-
to-peer communication among agents as well to-peer communication among agents as well various kinds of mediator, auction, discovery, various kinds of mediator, auction, discovery, service provider agents … and to see how well the service provider agents … and to see how well the semantic web works as the common knowledge semantic web works as the common knowledge infrastructure.infrastructure.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 88
TAGA: Travel Agent Game in Agentcities
http://taga.umbc.edu/
TechnologiesTechnologiesFIPA FIPA (JADE, April Agent Platform)(JADE, April Agent Platform)Semantic Web Semantic Web (RDF, OWL)(RDF, OWL)Web Web (SOAP,WSDL,DAML-S)(SOAP,WSDL,DAML-S)Internet Internet (Java Web Start )(Java Web Start )
FeaturesFeaturesOpen Market FrameworkOpen Market FrameworkAuction ServicesAuction ServicesOWL message contentOWL message contentOWL OntologiesOWL OntologiesGlobal Agent CommunityGlobal Agent Community
MotivationMotivationMarket dynamicsMarket dynamicsAuction theory (TAC)Auction theory (TAC)Semantic webSemantic webAgent collaboration (FIPA Agent collaboration (FIPA & Agentcities)& Agentcities)
Travel Agents
Auction Service Agent
Customer Agent
Bulletin BoardAgent
Market Oversight Agent
Request
Direct Buy
Report Direct Buy Transactions
BidBid
CFP
Report Auction Transactions
Report Travel Package
Report Contract
ProposalWeb Service
Agents
OntologiesOntologieshttp://taga.umbc.edu/ontologies/http://taga.umbc.edu/ontologies/ travel.owl travel.owl – travel concepts– travel concepts fipaowl.owl fipaowl.owl – FIPA content lang.– FIPA content lang. auction.owl auction.owl – auction services– auction services tagaql.owl tagaql.owl – query language– query language
FIPA platform infrastructure services, including directory facilitators enhanced to use OWL-S for service discovery
Owl for representation and reasoning
Owl for service
descriptions
Owl for negotiatio
n
Owl as a content languag
e
Owl for publishing
communicative acts
Owl for contract
enforcement
Owl for modeling trust
Owl for authorization policies
Owl for protocol
description
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 99
What we learnedWhat we learned OWL is a good KR language for a reasonably OWL is a good KR language for a reasonably
sophisticated MASsophisticated MAS Integrates well with FIPA standardsIntegrates well with FIPA standards
OWL made it easy to mix content from OWL made it easy to mix content from different ontologies unambiguouslydifferent ontologies unambiguously Supporting partial understanding & extensibilitySupporting partial understanding & extensibility
The use of OWL supported web integrationThe use of OWL supported web integration Using information published on web pages and Using information published on web pages and
integrating with web services via WSDL and SOAPintegrating with web services via WSDL and SOAP OWL has limitations: no rules, no default OWL has limitations: no rules, no default
reasoning, graph semantics, …reasoning, graph semantics, … Some of which are being addressedSome of which are being addressed
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1010
(2)(2) It’s policies all the way It’s policies all the way downdown
11 A robot may not injure a A robot may not injure a human being, or, through human being, or, through inaction, allow a human inaction, allow a human being to come to harm.being to come to harm.
22 A robot must obey the A robot must obey the orders given it by human orders given it by human beings except where such beings except where such orders would conflict with orders would conflict with the First Law.the First Law.
33 A robot must protect its A robot must protect its own existence as long as own existence as long as such protection does not such protection does not conflict with the First or conflict with the First or Second Law.Second Law.- Handbook of Robotics, 56th Edition, 2058 A.D.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1111
(2)(2) It’s policies all the way It’s policies all the way downdown
In Asimov’s world, the robots didn’t always strictly follow their policies Unlike traditional “hard coded” rules
like DB access control & OS file permissions
Autonomous agents need policies as “norms of behavior” to be followed to be good citizens
So, it’s natural to worry about … How agents governed by multiple
policies can resolve conflicts among them
How to deal with failure to follow policies – sanctions, reputation, etc.
Whether policy engineering will be any easier than software engineering
1 A robot may not injure a human being, or, through inaction, allow a human being to come to harm.
2 A robot must obey the orders given it by hu-man beings except where such orderswould conflict with the First Law.
3 A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
- Handbook of - Handbook of Robotics, 56th Edition, Robotics, 56th Edition, 2058 A.D.2058 A.D.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1212
Our ApproachOur ApproachPolicies are useful at virtually all levelsPolicies are useful at virtually all levels
OS, networking, data management, applicationsOS, networking, data management, applicationsDeclarative policies guide the behavior of Declarative policies guide the behavior of entities in open, distributed environmentsentities in open, distributed environments Positive & negative authorizations & obligationsPositive & negative authorizations & obligations Focused on domain actionsFocused on domain actions Policies are based on attributes of the action Policies are based on attributes of the action
(and its actor and target) and the general (and its actor and target) and the general context – not just on their context – not just on their identity identity of the actorof the actor
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1313
Rei Policy LanguageRei Policy LanguageDeveloped several versions of Rei, a Developed several versions of Rei, a policy specification language, encoded in policy specification language, encoded in (1) Prolog, (2) RDFS, (3) OWL(1) Prolog, (2) RDFS, (3) OWL
Used to model different kinds of policiesUsed to model different kinds of policies Authorization for servicesAuthorization for services Privacy in pervasive computing and the webPrivacy in pervasive computing and the web Conversations between agentsConversations between agents Team formation, collaboration & maintenanceTeam formation, collaboration & maintenance
The OWL grounding enables policies that The OWL grounding enables policies that reason over SW descriptions of actions, reason over SW descriptions of actions, agents, targets and contextagents, targets and context
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1414
Rei Policy LanguageRei Policy Language Rei is a declarative policy language for describing Rei is a declarative policy language for describing
policies over actionspolicies over actions Reasons over domain dependent informationReasons over domain dependent information
Currently represented in OWL + logical variablesCurrently represented in OWL + logical variables Based on deontic conceptsBased on deontic concepts
Permission, Prohibition, Obligation, DispensationPermission, Prohibition, Obligation, Dispensation Models speech actsModels speech acts
Delegation, Revocation, Request, CancelDelegation, Revocation, Request, Cancel Meta policiesMeta policies
Priority, modality preferencePriority, modality preference Policy engineering toolsPolicy engineering tools
Reasoner, IDE for Rei policies in EclipseReasoner, IDE for Rei policies in Eclipse
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1515
Applications – past, present & Applications – past, present & futurefuture
Coordinating access in Coordinating access in supply chain supply chain management systemmanagement system
Authorization policies in a Authorization policies in a pervasive pervasive computing environmentcomputing environment
Policies for team formation, Policies for team formation, collaboration, information flow in collaboration, information flow in multi-agent systemsmulti-agent systems
Security in Security in semantic web servicessemantic web servicesPrivacy and trust on the Privacy and trust on the InternetInternetPrivacy in Privacy in pervasive computing pervasive computing
environmentsenvironments
1999
2002
2003…
2004…
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1616
What we learnedWhat we learned Declarative policies can be used to model Declarative policies can be used to model
security, trust and privacy constraintssecurity, trust and privacy constraints Reasonably expressive policy languages Reasonably expressive policy languages
can be encoded on OWLcan be encoded on OWL This enables policies to depend on This enables policies to depend on
attributes and context information attributes and context information available on the semantic webavailable on the semantic web
Policies are applicable at almost every Policies are applicable at almost every level of the stack, from systems and level of the stack, from systems and networking to multiagent applications.networking to multiagent applications.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1717
(3)(3) A Love Triangle?A Love Triangle?
SemanticWeb
SoftwareAgents
PervasiveComputing
Even matches made in Heaven don’t always work out as planned.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1818
(3)(3) Pervasive Pervasive ComputingComputing
“The most profound technologies are those that disappear. They weave themselves into the fabric of everyday life until they are indistinguishable from it ” – Mark Weiser
Think: writing, central heating, electric lighting, water services, …
Not: taking your laptop to the beach, or immersing yourself into a virtual reality
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 1919
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2020
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2121
Pervasive environments for the Pervasive environments for the MilitaryMilitary
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2222
This is a challenging environmentThis is a challenging environment While devices are getting smaller, cheaper and
more powerful, they still have severe limitations. Battery, memory, computation, connection, bandwidthBattery, memory, computation, connection, bandwidth Each as limited sensors and perspectiveEach as limited sensors and perspective
The environment is inherently dynamic with The environment is inherently dynamic with serendipitous connections and unknown entitiesserendipitous connections and unknown entities
This makes security and trust importantThis makes security and trust important MANETS (mobile ad hoc networks) underlie MANETS (mobile ad hoc networks) underlie
pervasive infrastructures like Bluetoothpervasive infrastructures like Bluetooth It’s autonomous agents all the way downIt’s autonomous agents all the way down
Security and privacy is a special concernSecurity and privacy is a special concern Warfighters and agents must control how information Warfighters and agents must control how information
about them is collected and usedabout them is collected and used
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2323
Representing and reasoning about Representing and reasoning about contextcontext
CoBrACoBrA: a broker centric agent : a broker centric agent architecture for supporting pervasive architecture for supporting pervasive context-aware systemscontext-aware systems Using SW ontologies for context modeling Using SW ontologies for context modeling
and reasoning about devices, space, and reasoning about devices, space, time, people, preferences, meetings, etc.time, people, preferences, meetings, etc.
Using logical inference to interpret Using logical inference to interpret context and to detect and resolve context and to detect and resolve inconsistent knowledgeinconsistent knowledge
Allowing users to define policies Allowing users to define policies controlling how information about them controlling how information about them is used and sharedis used and shared
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2424
A Bird’s Eye View of CoBrAA Bird’s Eye View of CoBrA
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2525
SOUPA Ontology provides common SOUPA Ontology provides common vocabularyvocabulary
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2626
A Simple Spatial Model of UMBC
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2727
Where’s Harry?
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2828
Detecting Inconsistencies
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 2929
Powerful AI toolsPowerful AI tools DL reasoning over privacy policiesDL reasoning over privacy policies Abductive reasoningAbductive reasoning
Cobra uses a simple form of abductive reasoning to Cobra uses a simple form of abductive reasoning to keep track of assumptions underlying it’s conclusionskeep track of assumptions underlying it’s conclusionse.g.: A person’s at X if his cell phone’s at X assuming it’s not e.g.: A person’s at X if his cell phone’s at X assuming it’s not lostlost
This allows Cobra to find This allows Cobra to find consistentconsistent and and plausibleplausible explanations for the data it sees.explanations for the data it sees.
ArgumentationArgumentation Agents can question or challenge each others beliefs, Agents can question or challenge each others beliefs,
engaging in simple argumentation dialogsengaging in simple argumentation dialogse.g.: giving the assumptions and provenance underlying a beliefe.g.: giving the assumptions and provenance underlying a belief
An agent may choose not to adopt another’s belief An agent may choose not to adopt another’s belief based on the provenance of the databased on the provenance of the data
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3030
Privacy Protection in CoBrAPrivacy Protection in CoBrA Users define policies to permit or Users define policies to permit or
prohibit the sharing of their informationprohibit the sharing of their information Policies are provided by personal Policies are provided by personal
agents or published on web pagesagents or published on web pages and use the SOUPA ontologies as well and use the SOUPA ontologies as well
as other SW assertions (e.g., FOAF, as other SW assertions (e.g., FOAF, schedules)schedules)
The context broker follows user defined The context broker follows user defined policies when sharing information, policies when sharing information, unless contravened by higher policiesunless contravened by higher policies
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3131
The SOUPA Policy Ontology
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3232
Policy Reasoning Use CasePolicy Reasoning Use Case The speaker doesn’t want others to know The speaker doesn’t want others to know
the specific room that he’s in, but is willing the specific room that he’s in, but is willing for others to know he’s on campusfor others to know he’s on campus
He defines the following privacy policyHe defines the following privacy policy Share my location with a granularity >= “State”Share my location with a granularity >= “State”
The broker The broker isLocated(US) => Yes!isLocated(US) => Yes! isLocated(Maryland) => Yes!isLocated(Maryland) => Yes! isLocated(UMBC) => Uncertain..isLocated(UMBC) => Uncertain.. isLocated(ITE-RM210) => Uncertain..isLocated(ITE-RM210) => Uncertain..
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3333
What we learnedWhat we learned FIPA and OWL were good for integrating FIPA and OWL were good for integrating
disparate components, even on cell disparate components, even on cell phones!phones!
OWL made it easy to mix content from OWL made it easy to mix content from different ontologies unambiguouslydifferent ontologies unambiguously
OWL made it easy to take advantage of OWL made it easy to take advantage of information published in XML on the webinformation published in XML on the web e.g., foaf information, privacy policye.g., foaf information, privacy policy
Powerful AI tools add valuePowerful AI tools add value DL reasoning, abduction, argumentationDL reasoning, abduction, argumentation
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3434
(4)(4) TIVO for Mobile TIVO for Mobile ComputingComputing
A mobile computing vision and a problemA mobile computing vision and a problemDevices “broadcast” information and service Devices “broadcast” information and service descriptions via short-range RF (802.11, descriptions via short-range RF (802.11, Bluetooth, UWB, etc.)Bluetooth, UWB, etc.)
As people and their devices move, they can As people and their devices move, they can access this data, but only while it’s in rangeaccess this data, but only while it’s in range The data may be out of range when it’s neededThe data may be out of range when it’s needed
Devices must anticipate their information need Devices must anticipate their information need so they can cache data when it’s availableso they can cache data when it’s available Based on user model, preferences, schedule, Based on user model, preferences, schedule,
context, trust, …context, trust, … Compute a dynamic utility function to create a Compute a dynamic utility function to create a
“semantic” cache replacement algorithm“semantic” cache replacement algorithm
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3535
MoGATU’s distributed belief MoGATU’s distributed belief modelmodel
MoGATU is a data management module for MANETsMoGATU is a data management module for MANETs Devices send queries to peersDevices send queries to peers
Ask its vicinity for reputation of untrusted peers that responded Ask its vicinity for reputation of untrusted peers that responded -- trust a device if trusted before or if enough trusted peers trust -- trust a device if trusted before or if enough trusted peers trust itit
Use answers from (recommended to be) trusted Use answers from (recommended to be) trusted peers to determine answerpeers to determine answer
Update reputation/trust level for all responding Update reputation/trust level for all responding devicesdevices Trust level increases for devices giving what becomes final Trust level increases for devices giving what becomes final
answeranswer Trust level decreases for devices giving “wrong” answerTrust level decreases for devices giving “wrong” answer
Each devices builds Each devices builds a ring of trusta ring of trust……
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3636
A: Where is Bob?
C: I know where Bob is.
D: I know where Bob is.
B: I know where Bob is.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3737
A: D, where is Bob?A: C, where is Bob?A: B, where is Bob?
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3838
C: A, Bob is at work.
D:A, Bob is home.
B: A, Bob is home.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 3939
A:B: Bob at home,C: Bob at work,D: Bob at home
A: I have enoughtrust in D. Whatabout B and C?
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 4040
A: Do you trust C?
C: I always do.
D: I don’t.
B: I am not sure.
E: I don’t.F: I do.
A:I don’t care what C says.
I don’t know enough about B, but I trust D, E, and F. Together,
they don’t trust C, so won’t I.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 4141
A: Do you trust B?
C: I never do.
D: I am not sure.
B: I do.
E: I do.F: I am not sure.
A:I don’t care what B says.
I don’t trust C, but I trust D, E, and F. Together,
they trust B a little, so will I.
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 4242
A: I trust B and D,both say Bob is
home…
A:Increase trust in D.
A:Decrease trust in C.
A:Increase trust in B.
A:Bob is home!
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 4343
What we learnedWhat we learned OWL was a good language for capturing user OWL was a good language for capturing user
profiles and the simple BDI models we profiles and the simple BDI models we neededneeded
Any of several simple trust models increase Any of several simple trust models increase the accuracy of informationthe accuracy of information Designing a good trust model depends on Designing a good trust model depends on
the MANET assumptionsthe MANET assumptions As well as the level of cooperation and As well as the level of cooperation and
honestyhonesty Trading reputation information boosts the Trading reputation information boosts the
performance of the algorithmsperformance of the algorithms
4444
Knowledge DiscoveryKnowledge Discoveryin the Semantic Webin the Semantic Web(4)(4) SEMDIS NSF award ITR-IIS-0325464
U. Georgia, Sheth, Arpinar, Kochut, Miller NSF award ITR-IIS-0325172 UMBC, Joshi, Yesha, Finin
June 2004
ObjectiveDesign, prototype and evaluate a system supporting the discovery, indexing and querying of complex semantic relationships in the Semantic Web. The system maintains and utilizes trust and provenance information to enhance the relationship discovery.
ApproachKnowledge representation systems reason over sem-antic web content discovered on the web which is re-duced to triples that can be efficiently stored and pro-cessed in relational databases. Trust models and heuristics guide the formation of conclusions
Broader impactsTechniques and prototypes developed can be applied to a range of problems, including discovering new connections and relations in scientific information and homeland security.
SWETO is large ontology covering several test-bed domains. It is pop-ulated with 800K instances and 1.M relations extracted from heterogeneous Web sources. SWETO was developed using Semagix Freedom system.
An experimental algorithm has been developed to integrate and rank discovered relationships.
Reference foaf:Agent
rdf:Statementselects
Justification TrustBelief
Association
contains
foaf:Document
rdf:Resourcefoaf:page
DocumentRelation
xsd:real [0,1]AssociationConnectiveconfidence
connective
source
A “web of belief” model and associated ontology is used to represent, integrate, and evaluate conclusions drawn from the large volume of heterogeneous assertions found in the data.
http://lsdis.cs.uga.edu/Projects/SemDis http://semdis.umbc.edu/
A. Joshi
L. Ding
H. Chen
P. Kolari
F. Perich
Y. Yesha
J. Golbeck
J. Hendler
Kagal
sink
hub
source
island
Finin A. Joshi
Ding
Chen
Kagal
Perich
Golbeck’s Trust Network
DBLP Network
FOAF Network
A. Sheth
M. P. Singh
Y. Peng
6
15
1
28
T. Finin
mapTo
knows
knows
knows
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 4545
Need a slide or two here….Need a slide or two here….
4646Research support was provided by NSF, award NSF-ITR-IIS-0326460, PI Tim Finin, UMBC.
(5)(5) SPIRESPIRE Semantic Prototypes in Research EcoinfomaticsSemantic Prototypes in Research Ecoinfomatics
ApproachWe are building prototype tools and applications that demonstrate how semantic web technology supports infor-mation discovery, integration and sharing in scientific com-munities. The National Biological Information Infrastructure (NBII) and Invasive Species Forecasting System (ISFS) pro-vide requirements and serve as testbeds for our prototypes.
Significant Results• SWOOGLE - a search engine for the semantic web.• MoaM (Meal of a Meal) - Given a species list, infer a food web.• Photostuff - annotate regions of a picture with OWL.• SWOOP - the first ontology editor written specifically for OWL.• Ontologies for ecological interaction, and observation data.• Food web visualization and analysis tools that are driven by OWL
ontologies and instance data. • CRISIS CAT - an RDF based catalog of Invasive Species
resources in California. • Coordination with USGS, NASA, EPA, GBIF, and the
Intergovernmental, Interagency Cooperation on Ecoinformatics.
Broader Impacts• Enable knowledge from one community to be effectively
used by another.• Harness the power of the citizen scientist. (The majority of
invasives are discovered by amateurs.)• Integrate research and education in the classroom.
Research TeamUMBC ebiquity (Finin) UC Davis ICE (Quinn)UMBC GEST Center (Sachs) RMBL PEaCE (Martinez)UMD MINDSWAP (Hendler) NASA GSFC (Schnase)
The RMBL team expresses food webs in OWL using an ontology for eco-logical interaction they have constructed in coordination with other ecolo-gists. The OWL model drives the simulation and visualization.
Spatial distribution of exotic plants at the Cerro Grande fire site. The statistical techniques used to generate these maps do not take trophic data as input. Yet.
Swoogle is a crawler based search and retrieval system for semantic web doc-uments (SWDs) in RDF and OWL. It discovers SWDs and computes their metadata and relations, and stores them in an IR system. Users can search for ontologies or instance data, and hits are ranked according to our Ontology Rank algorithm.
Invasive species do more economic damage to the U.S. every year that all other natural disasters combined. Above: plants, animals, and a virus.
An ontology (found via Swoogle) is loaded into Photostuff to mark up regions of a field photograph.
The NBII California Information Node (CAIN), maintained by UC Davis, is a jumping off point to broader NBII deployment.
Coming Soon• ELVIS – an end to end application that starts with a location
and produces a model of its food web.• The Pond Project - a junior high school classroom activity to
monitor the health of local ecosystems. • Enhanced tools.
Spire is a distributed, interdisciplinary research project exploring how semantic web technology supports information discov-ery, integration, and sharing in scientific communities. We are building prototype tools and applications for inclusion in the National Biological Information Infrastructure (NBII), with a focus on the early detection and warning of invasive species.
Meal of a Meal (after Friend of a Friend). We know Fish 1 eats Plant 1. We then infer that Fish 1 may also eat the taxonomic siblings of Plant 1: Plants 2 and 3. Similarly, we infer that the taxonomic siblings of Fish 1 - Fishes 2 and 3 - may eat Plant 1.
UMBCUMBCAN HONORS UNIVERSITY IN MARYLAND
4747
Swoogle is a crawler based search & retrieval system for semantic web documents (SWDs) in RDF, Owl and DAML. It discovers SWDs and computes their metadata and relations, and stores them in an IR system.
Contributors include Tim Finin, Anupam Joshi, Yun Peng, R. Scott Cost, Jim Mayfield, Joel Sachs, Pavan Reddivari, Vishal Doshi, Rong Pan, Li Ding, and Drew Ogle. Partial research support was provided by DARPA contract F30602-00-0591 and by NSF by awards NSF-ITR-IIS-0326460 and NSF-ITR-IDM-0219649. 20 May 2004.
http://swoogle.umbc.edu/
Ontologydiscovery
Webinterface
DB SWDcrawler WeWe
bb
OntologyAnalyzer
OntologyAgentsOntology
AgentsOntologyAgentsOntology
Agents Ontologydiscovery Google
Apache/Tomcat
php, myAdmin
mySQL
Jena JenaIR
engine
SIRE
Webservices
Agentservices
cachedfiles
FocusedCrawler
APIs
Swoogle uses two kinds of crawlers to discover semantic web documents and several analysis agents to compute metadata and relations among documents and ontologies. Metadata is stored in a relational DBMS.
SWD RankA SWD’s rank is a function of its type (SWO/SWI) and the rank and types of the documents to which it’s related.
SWD Properties
SWOs
SWIs
HTMLdocuments
Images
CGI scripts
Audiofiles
Videofiles
The web, like Gaul, is divided into three parts: the regular web (e.g. HTML), Seman- tic Web Ontologies (SWOs), and Semantic Web Instance files (SWIs)
SWD = SWO + SWI
Binary: R(D1,D2)• IM: D1 owl:imports D2• IMstar: transitive closure of IM• EX: D1 extends D2 by defining classes or properties subsumed by D2’s
• PV: owl:priorVersion & subproperties• TM: D1 uses terms from D2• IN: D1 uses individual defined in D2• MAP: D1 maps some of its terms to D2’s• SIM: D1 & D2 are similar• EQ: D1 & D2 are identical• EQV: D1 & D2 have the same triplesTernary: R(D1,D2,D3) • MP3: D1 maps a term from D2 to D3 using owl:sameClass, etc.
SWD Relations
Language and level; encoding, number of triples, defined classes, defined properties, & defined individuals; type (SWO, SWI); form (RSS, FOAF, P3P, …); rank; weight; annotations; …
Swoogle puts documents into a character n-gram based IR engine to compute document similarity and do retrieval from queries
Swoogle has metadata on classes, properties and individuals from ~240,000 SWDs SWD IR Engine
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 4848
49UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
Uncertainty in Ontology Engineering: A Bayesian Perspective
Yun Peng, Zhongli Ding, Rong Pan
Department of Computer Science and Electrical engineering
University of Maryland Baltimore [email protected]
50UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
• Uncertainty in ontology engineering– In representing/modeling the domain
• Degree of inclusion• Overlap rather than inclusion• Uncertain information often available
– In reasoning• How close a description D is to its most specific subsumer?• Noisy data: leads to over generalization in subsumptions• Uncertain input:
– In mapping concepts from one ontology to another• Similarity between concepts may not be described logically• Mappings are hardly 1-to-1
• Uncertainty becomes more prevalent in web environment– One ontology may import other ontologies– Competing ontologies for the same or overlapped domain
Motivations
51UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
BN1
– OWL-BN translation• By a set of translation rules and procedures• Maintain OWL semantics• Ontology reasoning by probabilistic
inference in BN
Overview of The Approachonto1
P-onto1Probabilistic ontological information
Probabilistic ontological information
onto2
P-onto2
BN2Probabilistic annotation
OWL-BN translation
concept mapping
– Ontology mapping• A parsimonious set of links• Capture similarity between concepts by
joint distribution• Mapping as evidential reasoning
52UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
• Translated BN will preserve– Semantics of the original ontology– Encoded probability distributions among relevant variables
• Encoding probabilities in OWL ontologies– Define new OWL classes for prior and conditional
probabilities• Structural translation: a set of rules
– Class hierarchy: set theoretic approach• Each OWL class translated to a BN node• Arcs from super to subclass nodes
– Logical relations (equivalence, disjoint, union, intersection...)• Control nodes whose CPT realize logical relations
– Properties (ongoing)
OWL-BN Translation
53UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
Structural Translation• Set theoretic approach
– Each OWL class is considered a set of objects/instances– Each class is defined as a node in BN– An arc in BN goes from a superset to a subset– Consistent with OWL semantics
<owl:Class rdf:ID=“Human"> <rdfs:subclassOf rdf:resource="#Animal"> <rdfs:subclassOf rdf:resource="#Biped"> </owl:Class>
RDF Triples:
(Human rdf:type owl:Class)(Human rdfs:subClassOf Animal)(Human rdfs:subClassOf Biped)Translated to BN
54UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
Structural Translation• Logical relations
– Some can be encoded by CPT (e.g.. Man = Human∩Male)
– Others can be realized by adding control nodes
Man HumanWoman HumanHuman = Man WomanMan ∩ Woman = auxiliary node: Human_1Control nodes: Disjoint, Equivalent
55UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
Constructing CPT• Imported Probability information is not in the form of CPT• Assign initial CPT to the translated structure by some
default rules• Iteratively modify CPT to fit imported probabilities while
setting control nodes to true.– IPFP (Iterative Proportional Fitting Procedure)
To find Q(x) that fit Q(E1), … Q(Ek) to the given P(x)• Q0(x) = P(x); then repeat Qi(x) = Qi-1(x) Q(Ej)/ Qi-1(Ej) until
converging• Q (x) is an I-projection of P (x) on Q(E1), … Q(Ek) (minimizing
Kullback-Leibler distance to P)– Modified IPFP for BN
56UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
Example
57UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
• Formalize the notion of mapping• Mapping involving multiple concepts• Reasoning under ontology mapping
Ontology Mapping
58UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
• Simplest case: Map concept E1 in Onto1 to E2 in Onto2
– How similar between E1 and E2 – How to impose belief (distribution) of E1 to Onto2
• Cannot do it by simple Bayesian conditioningP(x| E1) = ΣE2 P(x| E2)P(E2 | E1) similarity(E1, E2)
– Onto1 and Onto2 have different probability space (Q and P)• Q(E1) ≠ P(E1)• New distribution, given E1 in Onto1: P*(x) ≠ΣP (x|E1)P(E1)
– similarity(E1, E2) also needs to be formalized
Formalize The Notion of Mapping
59UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
• Jeffrey’s rule– Conditioning cross prob. spaces– P*(x) =ΣP (x|E1)Q(E1)– P* is an I-projection of P (x) on Q(E1) (minimizing Kullback-
Leibler distance to P)– Update P to P* by applying Q(E1) as soft evidence in BN
• similarity(E1, E2)– Represented as joint prob. R(E1, E2) in another space R– Can be obtained by learning or from user
• Define map(E1, E2) = <E1, E2, BN1, BN2, R(E1, E2)>
Formalize The Notion of Mapping
60UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
Reasoning With map(E1, E2)
Q BN1
E1
P BN2
E2
R
E1 E2
Applying Q(E1) as soft evidence to update R to R* by Jeffrey’s rule
Using similarity(E1, E2): R*(E2) = R*(E1, E2)/R*(E1)
Applying R*(E2) as soft evidence to update P to P* by Jeffrey’s rule
61UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
• Multiple mappings– One node in BN1 can map to all nodes in BN2– Most mappings with little similarity– Which of them can be removed without affecting the overall
• Similarity measure: – Jaccard-coefficient: sim(E1, E2) = P(E1 E2)/R(E1 E2)– A generalization of subsumption– Remove those mappings with very small sim value
• Question: can we further remove other mappings – Utilizing knowledge in BN
Mapping Reduction
62UMBCUMBC
an Honors University in an Honors University in MarylandMaryland
• Current focuses: – OWL-BN translation: properties– Algorithms for ontology reasoning as probabilistic inference
over translated BN– Ontology mapping: mapping reduction
• Prototyping and experiments• Issues
– Complexity– How to get these probabilities
Current focuses
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 6363
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 6464
Marie’s slides go hereMarie’s slides go here
UMBCUMBCan Honors University in an Honors University in
MarylandMaryland 6565
backupbackup