Web of Belief: Modeling and utilizing Trust and Provenance as a part of the Semantic Web

Web of belief modeling and using trust and provenance in the semantic web l.jpg
1 / 50
0
0
1087 days ago, 457 views
PowerPoint PPT Presentation

Presentation Transcript

Slide 1

Web of Belief: Modeling and utilizing Trust and Provenance as a part of the Semantic Web Department of Computer Science and Electronic Engineering University of Maryland Baltimore County Li Ding Last upgraded: 3/10/2014

Slide 2

Outline Introduction Thesis Statement Research depiction Research arrange Preliminary Work The Web Of Belief Framework Evaluation Contributions to software engineering Thesis Schedule

Slide 3

Motivation The developing body of the Semantic Web Observations Information More Data encoded in Semantic Web dialect from numerous sources Various lingo Ontologies Information is overseen in two layer instrument as far as "Archive, Ontology, namespace, term" Physical layer: the web of semantic web reports Logical layer: the RDF chart More Semantic Web Tools Drive powers Industrial: Weblog, RSS, interpersonal organization sites Academic: look into undertakings

Slide 4

Motivation (cont'd) The Semantic Web has not accomplished a true "KB" Credibility & Consistency Facts are given by numerous sources w/o ensure Scalability Data is in tremendous sum Data is put away in an open and conveyed setting Utility Data is divided Bad URI Reference of asset & namespace in the Web of records Lack of relationship in the RDF diagram

Slide 5

Motivation (cont'd) Why provenance and trust Important ideas acquired from human world Multi-teach causes: social, epistemology, brain science The establishment of learning administration and induction Keys to believability appraisal and support Empirical heuristics, likewise the supplement technique, without area information to direct reason over validity. Unequivocal representation of legitimization follow. Great Heuristics to determine irregularity. Keys to viability and proficiency Knowledge can be overseen by Provenance other than Topic Trust lessens look intricacy

Slide 6

Thesis Statement This paper demonstrates that our Web Of Belief system, a provenance and trust mindful surmising structure, is basic and successful in determining answers with believability appraisal and avocation over the open, appropriated, and huge scale online learning base gave by the Semantic Web.

Slide 7

Research Description

Slide 8

General Description Goal: model and utilize provenance and trust in the SW to empower a sound "world KB". to empower trust layer in the Semantic Web Representation Encode provenance and trust Represent SW as KB Inference Hypothesis Test Trust arrange calculation Statement believability Justification Ontology Dictionary Term definition Class tree Management procurement & process information get to interface Inference space development

Slide 9

The Infrastructure of the Semantic Web Applications utilizes Reputation Service Web substance catalog seeks Directory/Digest Service SW Service discoverer SW Data discoverer digests Computing Services Data Service RDF record SW information benefit database (Web) archive

Slide 10

Assumptions Propositional information (actualities) Uncertain information with provenance Open and circulated information stockpiling

Slide 11

Relationship to Other Work Representation Logical formalisms of specialist model (AI) Truth hypothesis (Epistemology) Provenance Data get to Collaborative KB in open appropriated setting (DB) Learning operator models: learning and conduct (social learning & brain research) Inference Reason over indeterminate information (thinking)

Slide 12

Logical Formalisms Modal Logic - coherently formalize specialist Agent & activity (McCarthy,1969; Kanger-Porn-Lindahl) Agent & conviction and goal (Cohen, Levesque,1990) Agent & learning (Epistemic rationale) Agent & conviction (Doxastic rationale) Agent & commitment (Deontic rationale ) Other legitimate formalisms for trust and conviction Regan's formal structure for conviction and trust Josang's subjective rationale Abdul-Rahman's social trust display Jones and Firozabadi's coordinated rationale model of trust

Slide 13

Epistemology

Slide 14

Learning Agent models Objects to be scholarly Domain Trust Referral Trust Methods Histogram Feedback based

Slide 15

Reason over dubious learning Quantitative approach Certainty components - Mycin (Shortliffe, 1976) (out of date heuristic), like Fuzzy approach Possibility hypothesis: Fuzzy rationale (Zade, 1965;1976) Dempter-Shafer hypothesis (Dempster,1968; Shafer 1976) Subjective rationale Probabilistic hypothesis: Bayes Network (Pearl;1982) Qualitative approach Non-monotonic rationale

Slide 16

Two level information get to Datalog Logical level RDF information get to dialect (with provenance) Quads TriQL SPARQL Storage level Centralized triplestore Kowari Decentralized Search motor?

Slide 17

Example walkthrough Given a theory/inquiry in type of an accumulation of RDF articulations with or w/o factors Provenance where would I be able to discover them? where are the definitions for every term? Conviction( specialist, truth): Who said or affirmed so? Legitimize( actuality, truth): Trust Can I trust them and therefore utilize them in basic leadership How would I believe alternate operators

Slide 18

Representation Agent, information Provenance Trust Data get to Metadata RDF inquiry dialect Pattern extraction Transitive conclusion RDF stockpiling Inference Trust arrange derivation Credibility Probabilistic induction Scalability Domain channel Social channel Semantic Web Relationship to Other Work

Slide 19

Research Plan

Slide 20

Approach – the WOB system Representation WOB philosophy Model provenance and trust into the semantic web Explicit speak to the semantic web Represent SW as a KB as far as "specialist, proclamation, affiliation" Management Provenance mindful information get to dialect Social system extraction and combination Provenance and trust based learning base extension Inference Hypothesis validity appraisal Trust organize deduction Provenance and trust based conviction assessment Explicit avocation Ontology lexicon

Slide 21

Research Methodology Identify genuine issues with illustrations Approach issues Formalize issue Position issue in writing, and observe related work Find issues to be determined Design and execute arrangements Evaluation techniques Statistics Project application Survey

Slide 22

Artifacts to be delivered [Data] Web Of Belief Ontology [System] Swoogle metadata and hunt benefit [System] Ontology word reference [Data] Swoogle Statistics [System] SemDis Trust layer [Algorithm] Trust based conviction assessment [Algorithm] Trust based information development

Slide 23

Limitations Limited in online Semantic Web reports

Slide 24

Preliminary Work

Slide 25

WebOfBelief Ontology Entity: Document, Statement, Reference, Agent, Association Sub-classes: trust, conviction, defense, reliance Facets Confidence (restrictive likelihood) Connective (semantics) Provenance (Agent-record) Ownership/Authorship (Agent-Reference) conviction (Reference-Reference) legitimization (doc-doc) reliance Logical Formalisms

Slide 26

Web Of Belief (WOB) Conceptual Framework (v0.92) AssociationConnective xsd:real [0,1] certainty connective Association Justification Dependency Belief Trust foaf:Document Reference foaf:Agent chooses foaf:page dc:creator contains rdf:Resource rdf:Statement source wob:support wob:weaken wob:cause wob:imply wob:truthful wob:wise wob:knowledgeable wob:cooperative wob:believe wob:disbelieve wob:nonbelieve wob:imports wob:priorVersion

Slide 27

Data process benefit Support information get to dialect

Slide 28

Credibility Assessment Trust Network Inference Given a trust system, how to engender trust in order to assess trust between any two specialists Trust and provenance based explanation assessment Explicit Justification

Slide 29

Ontology lexicon?

Slide 30

Social system extraction and mapping

Slide 31

Application Trust based conviction assessment Trust and provenance mindful deduction Hypothesis testing and legitimization

Slide 32

Evaluation Validate determined trust relations: overview clients Validate execution of WOB derivation Compare comes about w or w/o trust & provenance Validate application utility: client report

Slide 33

Contributions A down to earth structure that makes the Semantic Web a KB The Web of Belief Ontology Semantic Web information process benefit Search and peruse instruments for SW Support of RDF information get to dialect? Surmising Judge data dependability The main work in portraying the Semantic Web trust and provenance mindful appropriated deduction

Slide 34

Dissertation plan Measures Size of information that could be handle Size of trust system Milestones Half-way completed

Slide 36

Trust Semantic Web P2P Possibility Theory Belief Theory the Semantic Web Representation Belief, trust Policy, govern SW administrations SW keen client Reputation benefit Inference Derive trust Belief combination Justification Inference Service SW benefit discoverer SW process SW information discoverer Heuristic pursuit Flexible inquiry SW client SW process Digest/Search Service SW information benefit Information insurance SW document SW Composer create Rich Information Text A blueprint of the Semantic Web

Slide 37

An illustration derivation Sorry I don't have it, Do you need US populace? Discover Washington Population disambiguation Which `Washington' do you mean? Affiliations Belief. Who knows what? RDF reference How to allude a portion of RDF diagram SW process Trusting provenance Credential based trust Reputation based trust Context/Role based trust Trusting substance accord setting sayings Sure! the accompanying SWDs/Agents realize that Trust arrange revelation Uncertainty and Precision Trust organize Here are the conviction/dependability for every special answer Justification Rule speak to speculation Justification instantiates control Oh Yeah! Answer X is tenable in light of the fact that it originates from government site Fill a RDF format Show me the total meaning of class X

Slide 38

Expected Contributions Framework Features for portray the Semantic

SPONSORS