Complex relationships for the semantic web

1
Complex relationships for the semantic web

• Amir Ali Khosravi
• 82702903

2

• The emphasis will shift from finding documents to
  finding facts, actionable information, and
  insights
• Discover relevant and interesting relationships
  amongst the entities that these documents describe

3
Introduction

• To take maximum advantage of awareness we need
  support meaningful information requests
• Current ontological represetational schemes
  represent knowledge as a hierarchical taxonomy of
  concepts and relationships such as
  is-a/role-of,instance of/member-of and part of
• Support limited complexity
• InfoQuilt extend support for semantics by
  supporting computations involving lateral, user
  defined relationships
• Relationships across domains, may not
  necessareily hierarchical in nature

4
Introduction

• Does nuclear Testing cause Earthquakes?
• Natural-Disasters.Earthquake, Nuclear-Weapons.Nucl
  ear-Testing
• The meaning of cause could be based on the
  proximity in time and distance between two events

5
Classification of relationships

• Content independent relationships two documents
  may be relatedd to each other by virtue of them
  being stored on the same server or file system,
  or relationship between a document and its date
  of modification

6

• Content Dependent relationships either the
  information content they refer to or based on
  some representation of it thereof
• Direct Content dependent relationships relation
  between two entities being mentioned in the same
  pragraph
• Content descriptive relationships the fact that
  X is CEO of a company Y is computed based on the
  existence of an ontlogy (intra-domain
  relationships and inter-domain relationships

7
Content descriptive relationships

• Direct Semantic relationships Intel
  is-a-competitior-of Motorola
• Complex Transitive RelationshipsRemzi and Dick
  linked to the same teroorist organization
• Inter-domain Multi-ontology Relationships

8
Representation of relationships

• Arity
• Cardinality
• Direct vs. transitive relationships
• Crisp vs. fuzzy
• Properties vs. relations
• Structural composition

9

• Deficiency shared by most of these languages is
  the absence of a mechanism that can model complex
  operators
• Earthquake causing tsunami Temporal and spatial
  proximity
• InfoQuilts supports such operators by allowing
  the use of user defined functions as operators
• The system particularly needs to extract
  domain-specific or contextually relevant metadata
  from all the data sources
• Knowledge of the characteristics of the domains
  of interest gives us the ability to optimize the
  processing of information requests

10

• A form of knowledge discovery HAND
• Domain modeling uses ontologies, domain rules,
  functional dependency, support of user defined
  inter-ontological relationships
• Rich and powerful querying mechanism IScape
• User defined functions as complex operators
• Ability to handle heterogeneous content
• Information request processing utilizing domain
  and resource characteristics

11
Domain Modeling

• InfoQuilt uses ontologies to model the domains

12

• Attributes
• Earthquake(latitude, longitude, region,
  eventDate, Description, damagePhoto,
  numberOfDeaths, magnitude)
• Domain Rules
• latitude gt -90 , latitude lt 90
• Earthquake(latitude, longitude, region,
  eventDate, Description, damagePhoto,
  numberOfDeaths, magnitude,
• latitude gt -90 , latitude lt 90
• longitude gt -180, longitude lt 180)

13

• Functional Dependencies(FD)
• is used to retrieve information (attribute
  values) that is missing from a resource by using
  another resource
• testSite -gt latitude longitude
• Earthquake(latitude, longitude, region,
  eventDate, Description, damagePhoto,
  numberOfDeaths, magnitude,
• latitude gt -90 , latitude lt 90
• longitude gt -180, longitude lt 180,
• testSite -gt latitude longitude )

14
Interontological relationships

• We can say that some nuclear test could have
  caused an earthquake if we see that the
  earthquake occurred some time after the nuclear
  test was conducted and in nearby region
• NuclearTest Causes Earthquake
• lt dateDifference (NuclearTest.eventDate,
  Earthquake.eventDate) lt 30 AND distance(
  NuclearTest.latitude, NuclearTest.longitude,
  Earthquake.latitude, Earthquake.longitude) lt 10000

15
Operations

• Find all earthquakes with epicenter in a 5000
  mile radius area of the location at latitude
  60.790 North and longitude 97.570 East
• Another important advantage the system can
  support complex post-processing of data
• To be able to dynamically and easily add new
  operations as well as update and delete existing
  ones, InfoQuilt maintains Function Store

16
Information Scapes (IScapes)

• query generally explicitly specifies the exact
  sources, how data from these sources should be
  integrated, it does not understand what the user
  is asking
• IScape, can understand what the user is inquiring
  by embedding semantic information

17

• a computing paradigm that allows users to query
  and analyze the data available from a diverse
  autonomous sources, gain better understanding of
  the domains and their interactions as well as
  discover and study relationships
• Find all earthquakes with epicenter in a 5000
  mile radius area of the location at latitude
  60.790 North and longitude 97.570 East and find
  all tsunamis that they might have caused

18

• The preset constraint and runtime configurable
  constraint , similar to WHERE
• IScape builder to construct and execute IScapes
  and analyze results

19
Human Assisted Knowledge Discovery(HAND)

• Transitive relationships Earthquake causes
  Tsunami, Tsunami affects Enviornment,
  Earthquake affects Enviornment
• IScape 1 When was the earliest recorded nuclear
  test conducted
• IScape 2 Find the total number of earthquakes
  with a magnitude 5.8 or higher on the Richter
  scale per year starting from year 1990
• IScape 3 Find the average number of earthquakes
  per year with a magnitude 5.8 or higher on the
  Richter scale for the period 1900-1949 and for
  the period 1950-present

20

• IScape 4 For each group of earthquakes with
  magnitudes in the ranges 5.8-6, 6-7, 7-8, 8-9,
  and magnitudes higher than 9 on the richter scale
  starting from year 1900, find the number of
  earthquakes
• IScape 5 Find nuclear tests conducted after
  january 1,1950 and find any earthquakes that
  occurred not later than a certain number of days
  after the test and such that its epicenter was
  located no farther than a certain distance from
  the test site

21
InfoQuil runtime architecture

• Multi-agent information brokering architecture at
  runtime

22
Steps of precessing of an IScape

• User agent sends IScape to the broker agent for
  processing
• Broker agent sends the IScape to the planning
  agent
• Planning agent creates an execution plan for the
  IScape
• Planning agent returns the plan to the broker
  Agent
• Broker agent sends the plan to the correlation
  agent for processing
• Correlation agent starts executing the plan
• Correlation agent returns the final result to the
  broker agent
• Broker agent forwards the result to the user
  agent
• Supports easy dynamic addition and removal of
  resources

23
Information Extraction and integration

• Extract it as needed Vs. extract it offline and
  maintain it in a local database
• Important metrics for Information extraction
  recall , precision
• Information integration systems usually provide a
  uniform means of representing the information
  from multiple sources
• It is done offline and for sources with
  relatively static data
• METIS toolkit is used to create a single
  repository of information for a domain

24
METIS architecture
25
Resource Modeling

• Resource Attributes
• SignificantEarthquakesDB( eventDate, Description,
  region, magnitude, latitude, longitude,
  numberofDeaths, damagePhoto)
• EarthquakesAfter1990( eventDate, region,
  magnitude, numberOfDeaths, damagePhoto )
• Binding Patterns (BP)
• A set of attributes that the system must be able
  to suply values for in order to query the
  resource
• fromCity, fromState, toCity, toState,
  departureDate

26

• Data Characteristic (DC) Rules
• AirTranAirways (airlinecompany, flightNumber,
  fromCity, fromState, toState, DepartureDate,
  fare, departureTime, arrivalTime, dc
  airlineCompany AirTran Airways, fromCity,
  fromState, toCity, toState, departureDate )
• Find all the flights operated by Delta Airlines
  from Boston, MA to Los Angeles, CA on February
  19,2001

27

• Local Completeness (LC) Rules
• All the information for the subset
• AirTranAirways (airlinecompany, flightNumber,
  fromCity, fromState, toState, DepartureDate,
  fare, departureTime, arrivalTime, dc
  airlineCompany AirTran Airways, lc
  airlineCompany AirTran Airways)

28
Planning and optimization

• Two ontologies NuclearTest, Earthquake
• Specifications of the information sources
• NuclearTestsDB( testSite, explosiveYield,
  bodyWaveMagnitude, testType, eventDate,
  conductedBy, dcbodyWaveMagnitudegt 3, dc
  eventDate gt January 1, 1985 )
• NuclearTestSites (testSite, latitude, longitude
  )
• SignificantEarthquakeDB( eventDate, description,
  region, magnitude, latitude, longitude,
  numberOfDeaths, damagePhoto, dc eventDategt
  january 1, 1970 )

29
Planning and optimization

• Planning agent uses the following rules to select
  the resources that are relevant
• Locally complete Sources
• Non Locally Complete Sources
• Binding Patterns
• Associate resource to supply values for missing
  attributes

30

• NuclearTestsDB has two missing attributes
  latitude nad longitude, planner uses FD testSite
  -gt latitude longitude using NuclearTestSites as
  an associate resource, function testSiteEquals
  from the function store is used

31
IScape Execution and monitoring

• User agent passes the IScape to the Broker agent.
  
• The broker agent starts the processing of the
  IScape by coordinating other agents,
• first asks planning agent to create an execution
  plan planning agent interacts with the knowledge
  agent to access information about domains, in
  IScape and resources
• Creates an execution plan
• Sends it back to the broker
• Broker sends to correlation agent
• Final result returned to broker and user agent

32
IScape processing monitor

• Execution of the plan in the correlation agent is
  multi-threaded and parallel

33
Knowledge Builder (KB)

• Create specifications of domains, inter-domain
  relationships, operations and the available
  information sources

34
IScape Builder (IB)

• Provides a graphical interface to create and
  execute IScapes
• Step1 select the ontologies that he/she wants,
  selection of the relationships

35

• Step2 specify functions that operate on
  attributes
• step3 specify the conditions that make up the
  structure of IScape
• Step 4 specify additional constraints on the way
  the result of the IScape query should be groupd
• Step5 specify the runtime projection parameters

36
(No Transcript)
37
References

• Amith Sheth, Sanjeev Thacker and Shuchi Patel,
  complex Relationships and knowledge Discovery
  support in the InfoQuilt system
• Amith Sheth, I.Budak Arpinar, and Vipul Kashyap,
  Relatiosships at the heart of semantic Web
  Modeling, Discovering, and Exploiting complex
  semantic relationships
• R. J. Bayardo Jr., W. Bohrer, R. Brice, et al.
  InfoSleuth Agent-Based Semantic
• Integration of Information in Open and Dynamic
  Environments. In SIGMOD-97,pp. 195-206, Tucson,
  AZ, USA, May 1997.
• C. Bertram. InfoQuilt Semantic Correlation of
  Heterogeneous Distributed
• Assets. Masters Thesis, Computer Science
  Department, University of Georgia,1998.

38
References

• Alexandria Digital Earth Prototype.
  http//www.alexandria.ucsb.edu/
• S. Adali and R. Emery. A uniform framework for
  integrating knowledge in heterogeneous systems.
  Proceedings of the Eleventh IEEE International
  Conference of Data Engineering (March 1995).
• Y. Arens, C. Hsu and C. A. Knoblock. Query
  processing in the SIMS
• information mediator. In Austin Tate, editor,
  Advanced Planning Technology.The AAAI Press,
  Menlo Park, CA, 1996.
• J. Ambite, and C. Knoblock. Planning by
  Rewriting Efficiently generating highquality
  plans. Proceedings of the 14th National
  Conference on Artificial Intelligence,
  Providence, RI, 1997.
• Y. Arens, C. A. Knoblock, and W. Shen. Query
  reformulation for dynamic information
  integration. Journal of Intelligent Information
  Systems, Vol. 6, pp. 99-130, 1996.

39
Any Question