Knowledge Base Grid Power the Internet By Intelligence

1
Knowledge Base GridPower the Internet By
Intelligence

• Zhaohui Wu PhD / Professor
• Grid Computing Lab of Zhejiang University

2
Outline

• Motivation
• Generic Architecture
• Knowledge Representation for
• Grid/Internet/Web
• Knowledge Protocol and Inference Service
• Summary
• Future Work

3
Part IMotivation
4
The Global Information Problem

• Internet has provided us a large-scale, universal
  and global information resource space
• But we have failed to organize, manage and
  utilize it well enough.
• We face so many problems and challenges
• Getting drowned in the tremendous web pages,
• Increasing difficulty to discover, organize, and
  retrieve the available information resources
  (search engine help us so little),
• Poor organization and integration of web
  applications and resource.

5
New Requirements

• In the other hand we have hoped so much for the
  future Internet
• Universal access into the web
• Single sign-on
• Easy to discovery and easy to sharing of
  information /knowledge resources in a worldwide
  way
• Autonomous discovery and semantic integration of
  web applications
• On-demand intelligent service satisfying our
  personal need and expectations
• Web should serve us not let so many things to be
  done by us.
• Intelligent information retrieval
• it should retrieve the interesting information
  satisfy personal expectations autonomously, not
  let human do clicks and clicks again. We hate the
  result returned by search engine
• Web Planning
• When we launch a task , it should be fulfilled
  intelligently without further human efforts.

6
Knowledge and Semantic Come of Age

• Internet / Web need intelligence
• We fail to manage it , so we hope machine do
  much more for us. We should make web/internet
  machine-understandable.
• HTTP / HTML /XML are not enough, for they
  lack semantic
• Knowledge enable web intelligence
• Not only we want to do query ,but also we want
  the machine to do inference for us.
• Database is not enough for web/Internet,
  for is has no ability of reasoning
• Semantic Web
• We could view Semantic Web as a large-scale
  virtual Knowledge Base in which knowledge has
  been encoded into web document by DAMLOIL /OWL
  /RuleML ,and the knowledge could be consumed by
  intelligent agent.
• Semantic Web Service
• Semantic / Knowledge enable intelligent
  discovery and autonomous integration of web
  services.
• SOAP / WSDL is not enough

7
Artificial Intelligence Get a Renaissance

• AI has provide us a great of theories and models
  ,such as model theory semantic?description logic
  ?ontology and so on
• AI has also provide us a great of methods and
  experience to represent knowledge, such as
  semantic net?frame-based system and so on
• Open Rule-based System or Open Knowledge Base or
  Open Expert System on the web

8
Our Goal and Solution

• The ultimate goal is bringing the Internet (Grid
  and Web) into its full potential by Artificial
  Intelligence.
• We design the enabling architecture and
  technologies to support machine facilitated
  global knowledge exchange, sharing and reuse .
• We want to create a knowledgeware Internet in
  which information/knowledge have been represented
  uniformly and well-organized into the web-KBs,
  and the information /knowledge resources can be
  wide-spread shared and coordinated used to
  provide on-demand and intelligent service
  satisfying personal needs and expectations.

9
Part IIGeneric Architecture
10
Terminology and Abbreviations

• KB-Grid Knowledge Base Grid
• VOKB Virtual Open Knowledge Base
• The VOKB is a knowledgware Internet entities
  providing coordinated inference service in which
  multiple WebKBs participate for distributed
  problem solving and planning.
• KS Knowledge Source
• A KS is the source of knowledge. These knowledge
  sources include Concepts, Facts, Procedures and
  Meta-Knowledge sources. They are the basic
  elements of the knowledge base.
• Inference Service
• Inference Service is a knowledge service, which
  provides high-level intelligent services such as
  backward-chaining inference service, subsumption
  and classification inference services etc.

11
Whats Knowledge Base Grid

• The KB-Grid enables a knowledgeware Internet in
  which knowledge have been represented uniformly
  and well-organized into the knowledge bases
• The knowledge base resources can be wide-spread
  shared and coordinated used to provide on-demand
  and intelligent services satisfying personal
  needs and expectations.
• The knowledge protocols enable such a coordinated
  sharing of KB resources.
• The KB-Grid focus on constructing large-scale and
  virtual open KBs (VOKB) which support global
  knowledge acquisition , management, discovery,
  sharing and reuse in the Internet environment.

12
The Core Components of KB-Grid

• Decentralized generic knowledge base resources
• Large-scale VOKBs (Virtual open knowledge base
  recourses )
• Hierarchical KB-GIIS for collecting, indexing the
  information about the knowledge base resource
• Semantic Browser as uniformed client of KB-Grid
  Semantic Browser owns reasoning ability

13
The KB-Grid Big Picture
14
The Knowledge Base Resource

• Web KB resources
• Semantic Web will bring us great of KB
  resources, KB-Grid should provide method to
  manage and organize them , and we could envision
  great of knowledge-intensive systems in
  Grid/Internet/Web
• Each KB resources could have many Knowledge
  sources
• Those KS include Ontology (taxonomy) KS, Horn
  Rules KS, Cases KS.
• KB resource support some inference services
• Subsumption and classification inference service
  for taxonomy
• Backward-chaining and Forward-chaining inference
  service for horn rules
• Analogy inference service for Case knowledge

15
VOKB Virtual Open KB

• VOKB is some large-scale Web-KB providing
  coordinated inference services in which multiple
  KBs participate for distributed problem solving.
• VOKB will become popular in Semantic Web
• Because, most of the tasks will involve multiple
  knowledge providers. For example , in a touring
  schedule planning scenario , the schedule task
  will involve the map information provider, hotel
  information provider, the traffic information
  provider, and so on

16
KB-GIIS

• KBGIIS KB-Grid Information Index Service
• KB-GIIS plays a role of KB resource aggregate
  directory
• Thats the search engine for KB resource
• The KB-GIIS is configured so that the KB-GRP can
  be used for both registration and enquiry.
• In registration, a knowledge base resource
  registers with a KB-GIIS explicitly. (like a web
  site register itself to search engine)
• In the case of invitation, a knowledge base
  resource is asked to join by the KB-GIIS. (like a
  search engine spider collect web sites
  information into its databases)
• Note
• The LDAP model underlying GIIS of computing grid
  is not enough, we need more knowledgeware data
  model

17
Semantic Browser

• The Client for KB-Grid (also for Semantic Web)
• The most important characteristic of it is it has
  some embedded inference module
• Give a web ontology such as a product category of
  a online-shop, it will facilitate the information
  location, collection and classification

18
Three Key Issues on KB-GRID

• The KB-Grid infrostructure should take
  into consideration of the following three key
  issues
• Standard knowledge representation supporting well
  organization of information maybe database is
  not enough
• Standard knowledge protocols supporting semantic
  interoperation dynamic clustering and
  autonomous integration of decentralized
  information/knowledge resources
• On-demand intelligent service satisfying personal
  need and expectations.

19
Why Knowledge Representation?

• Knowledge and Semantic enable intelligence
• HTML , XML, Database are not enough
• Web resource description
• create meta-data for Web (thats why RDF
  originated form meat-data activity of W3C)
• Logic and inference in Web
• supporting construct large-scale web-KBs
• The most important is the standardization.

20
Why knowledge protocol

• Distributed control and semantic interoperability
  
• Eliminate the semantic difference among
  decentralized web applications and support
  semantic integration of those applications
• Knowledge protocols are more complex than
  computing grid protocol
• Both LDAP and SOAP are not enough for knowledge
  level communication
• The control process of communication become more
  complex, not just request-and-then-response

21
Why Intelligent Service

• In the future, the services available to us not
  only are on demand - serve us whenever we need,
  but also should be intelligent.
• The intelligence means that we should automate
  knowledge service discovery, composition, and
  semantic interoperation.
• These intelligent services should satisfy our
  personal needs and expectations without further
  human efforts.

22
Part IIIKnowledge Representation for
Grid/Internet/Web
23
Two purposes of Knowledge Representation for
Semantic Web

• Web resources description
• support more intelligent way of web resource
  locating, collecting and classification
• Support construct large-scale WebKBs
• for Network inference and Web Logic

24
The language for the Web

• HTMLnothing but free or semi-free text, so ugly
  !
• XML Structure of Content, but most important
  its the standard method for data interchange in
  Internet, but lack semantic.
• RDF/RDFS basic model for web resource
  description but they are the base for high level
  knowledge representation.
• DAMLOIL (DAPPA and EC)thats the description
  logic for web, concept and concept relation
  description, Terminology and Assertions
• OWL (W3C) the same as DAMLOIL, but its the
  standard
• RuleML (DFKI) thats the horn logic for web,
  publish rules onto the web,

25
Whats in the link of semantic web

• Essentially speaking, link stands for the
  relationship between two entities.
• For example, the hyper links
• lta hrefhttp//www.zju.edu.cngtZhejiang
  Univer.lt/agt
• stands for the relationship between the
  string Zhejiang Unver. and the network
  accessible web resource identified by
  http//www.zju.edu.cn.
• There are essential differences in the link of
  RDF.

26
The link in the RDF
27
Whats the difference?

• The key issue here is the isolation of the
  primitives for network languages such as HTML,
  XML, RDF/RDFS, DAMLOIL, etc.
• The primitives are those things that the
  interpreter is programmed in advance to
  understand, and that are not usually represented
  in the network language itself
• e.g., the primitive ltagt of HTML is interpreted by
  an interpreter embedded in the browser. When a
  click action has been performed, the browser
  knows how to direct the user to proper pages.

28
The four levels of link
Linguist Level
Concept Level (The concepts and roles)
Logic Level (AND OR SUBSET)
Physical Level ( Hyper Link, XML Link)
29
We Present the KML for KB-GRID

• A hybrid knowledge markup language (KML) cover
  the range of description logic, horn rules and
  cases-base reasoning
• Consisting of four components
• T-Box for concept definition,
• A-box for making assertion about resources,
• R-Box for rules
• C-Box for cases.
• The rules and cases are also web resources which
  we could make assertion about them by ABox
• Combing horn rules and description logic.
• Case markup and combining rule-based and
  case-based reasoning in the web
• KML support network inference and hybrid
  reasoning in KB-Grid

30
The requirements for hybrid Reasoning in KB-Grid

• When ask KB-Grid where and how we could buy some
  commodity, we should combine product ontology and
  a financial ontology which is used to pay money
• In e-business settings, we should combine some
  business rules (some forms like horn rules) with
  a product ontology and/or a enterprise ontology
  to finish such kind of tasks
• In a bioinformatics e-science scenario, we maybe
  need to integrate enzyme taxonomy with some
  chemosynthesis rules to describe a biochemical
  process.
• When providing case-based reasoning service such
  as a kind of semantic web service which provides
  on-line financial report analysis for enterprise
  by a case library , we need not only a taxonomy
  to provide semantic index about cases, but also
  some rules to assist selecting proper cases.

31
We need do much more on KR

• Hybrid reasoning
• Coordinated inference
• Knowledge Interchange
• Standardization

32
Part IVKnowledge Protocols and Inference Service
33
Knowledge Level Communication

• knowledge level communication support
• Semantic integration of applications
• Decentralized control
• Semantic interoperability
• Coordinated Inference
• Knowledge protocols is more complex
• The message format for communication should
  contain the knowledge interchanged
• The control process is more complex here
• OGSA maybe need knowledge protocols for
  application intelligent discovery and autonomous
  integration in large-scale environment such as
  Grid/Internet/Web.

34
Network Inference

• Network Inference has found a way for
  applications to understand each other.
• "Semantic differences, remain the primary
  roadblock to smooth application integration, one
  which Web Services alone won't overcome. Until
  someone finds a way for applications to
  understand each other, the effect of Web services
  technology will be fairly limited. When I pass
  customer data across the Web in a certain
  format using a Web Services interface, the
  receiving program has to know what that format
  is. You have to agree on what the business
  objects look like.  And no one has come up with a
  feasible way to work that out yet -- not Oracle,
  and not its competitors...
• ---
  Oracle Chairman and CEO Larry Ellison

35
The Protocols in KB-Grid

• KB-GIP Knowledge Base Grid Information Protocol
• The KB-GIP is used for inquiring about the
  information about the KB or VOKB in the Grid by
  the KB-GIS
• KB-GRP Knowledge Base Grid Registration Protocol
• The KB-GRP is responsible for knowledge base
  resource registration and enquiring the
  patricians of coordinated sharing.
• KB-GQMP Knowledge Base Grid Query and
  Manipulation Protocol
• The KB-GQMP answers for knowledge query and
  manipulation such as knowledge browsing and
  coordinated inference.

36
The Service in KB-GRID

• Basic Services Resource Management
• The basic services answer for the management and
  organization of knowledge, and coordinated
  sharing , reuse , dynamic integration of
  knowledge base resources
• Inference Service High Level Knowledge Services
• Inference Service is a knowledge service, which
  provides high-level intelligent services such as
  knowledge query service, backward-chaining
  inference service, forward-chaining inference
  service, subsumption and classification inference
  services etc.

37
Basic Service in KB-GRID

• KB-GIS KB-Grid Information Service
• KB-GIS is designed to support the initial
  discovery and ongoing retrieval of the
  characteristics and information of knowledge base
  resource.
• KB-GIIS KB Grid Index Information Service
• A KB-GIIS plays a role of knowledge service
  aggregate directory that uses KB-GRP and KB-GIP
  to obtain information (from the KB-GIS of a
  knowledge base resource) about a set of
  intelligent entities, and then replies to queries
  concerning those entities.
• KB-GAMS Knowledge Acquisition and Management
  Service The KB-GAMS provide a worldwide way for
  knowledge acquisition and management in the
  KB-Grid.

38
Inference Service in KB-GRID

• Knowledge Query Service
• Correlative Semantic Browsing Service
• Subsumption and Classification Inference Service
• Forward and backward-chaining inference service
• And so on

39
Inference Service on (Web) Ontology

• Completion inference service
• Logical consequences of assertions about
  individuals and descriptions of concepts are
  computed. These include Inheritance, Propagation,
  Contradiction detection and Incoherent concept
  detection
• Subsumption and Classification inference
  service
• Given a new concept and a knowledge base
  resource that contains concepts about some domain
  of discourse, the inference service should return
  all the concept subsumed by the new concept or
  place the new concept into proper position in the
  virtual WebKB and construct right association
  with other concepts.

40
Inference Service on(Web) Rules

• Forward Chaining Inference Service
• Given a new fact, the knowledge base
  resource in Semantic Web should derive some of
  the facts implied by the conjunction of the
  knowledge base and the new fact, and then more
  inferences can be made on the derived
  conclusions. This is called forward-chaining
  inference on KML. If the coordinated inference
  ability has been turned on, it should also route
  the inference request to other correlative KBs
  and return all derived facts to user
  transparently.
• Backward Chaining Inference Service
• Alternatively, we can start with something we
  want to prove, find implication sentences that
  would allow us to conclude it and then attempt to
  establish their premises in turn. Given a goal
  and some optional facts to the knowledge base
  resource, the inference service should answer us
  the true or false about the goal and give us an
  explanation about the inference process. This is
  called Backward Chaining inference service. In
  such a setting, we may also need a coordinated
  inference process which multiple knowledge
  resources participate to achieve a goal.

41
Part VSummary and Future Work
42
What we have done?

• Global Information Problems and new Requirements
• The Generic Architecture for KB-Grid
• The KB-Grid enables a knowledgeware Internet in
  which knowledge have been represented uniformly
  and well-organized into the knowledge bases and
  the knowledge base resources can be wide-spread
  shared and coordinated used to provide on-demand
  and intelligent services satisfying personal
  needs and expectations.
• Knowledge Representation for KB-Grid
• KML support hybrid reasoning and coordinated
  inference

43
What we will do?

• Make more deep research on Knowledge
  Representation for Grid and Semantic Web
• Network Inference Internet-Oriented inference
• Knowledge Protocols
• Semantic Web Service

44
Related Work

• The Knowledge Grid effort in Chinese Academy of
  Sciences
• Semantic Web Service in DARPA (DAML_S)
• Semantic Grid Work Group in GGF

45
Thanks!