42
International Technology Alliance In Network & Information Sciences Dr David Mott IBM UK Discussion with Prof Harold Boley 3 rd Dec 2009

International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

Embed Size (px)

Citation preview

Page 1: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

International Technology AllianceIn Network & Information Sciences

International Technology AllianceIn Network & Information Sciences

Dr David MottIBM UK

Dr David MottIBM UK

Discussion with Prof Harold Boley3rd Dec 2009

Discussion with Prof Harold Boley3rd Dec 2009

Page 2: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[2]

Suggested Agenda

• The ITA programme

• The “RIF enterprise”

• Your work

• How can we help?

Page 3: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[3]

International Technology Alliance in Network and Information Sciences

Page 4: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[4]

ITA Programme

• Focus– Enabling coalition operations over collaborative network centric systems

• Technical Areas– TA1 Network Theory

– TA2 Security across a system of systems

– TA3 Sensor information processing and delivery

– TA4 Distributed Network Enabled Cognition

• Aspects– From 2006 to 2011

– 1st research defence collaboration between US (ARL) and UK (MOD)

– Must be fundamental research published in open forum

– Joint academia and industry, US and UK

– http://usukita.org/

Page 5: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

U.S.Gov.

Industry

Academia

U.K.Gov.

INDUSTRY9. BBNT Solutions LLC

10.The Boeing Corporation

11.Honeywell Aerospace Electronic Systems

12. IBM Research

13.Klein Associates

ACADEMIA1. Carnegie Mellon University

2. City University of New York

3. Columbia University

4. Pennsylvania State University

5. Rensselaer Polytechnic Institute

6. University of California Los Angeles

7. University of Maryland

8. University of Massachusetts

INDUSTRY 8. IBM UK

9. LogicalCMG

10.Roke Manor Research Ltd.

11.Systems Engineering& Assessment Ltd.

ACADEMIA1. Cranfield University, Royal Military

College of Science, Shrivenham

2. Imperial College, London

3. Royal Holloway University of London

4. University of Aberdeen

5. University of Cambridge

6. University of Southampton

7. University of York

7

10

6

42

853

1

9

1312

11

123

4

5

6

7

8 91011

ITA Team Overview

Page 6: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[6]

Plan Representation

AnalyzingCommunication Patterns

Interpretation of human activity

Information Flow Analysis

Interpretation of human activity

Computer Mediated Interactions

Cultural Modelling - Planning and Intent

Battlefield IIUS/UK

Technical Area 4 – Network Enabled Cognition

Agent Support for ad-hoc Adaptive Teamwork

Ontologies and Semantic

Representations

Page 7: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[7]

• How can a shared understanding of a plan or other “artifact” be obtained?

– How can plan details be communicated and understood across different planners?

– How can a Commander describe his intent and rationale to the planners and operations staff?

ITA Project 12 Task 3 research focus

Page 8: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[8]

Collaborative Planning Model

Page 9: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[9]

Plan representation

Visualisation

CNL Rationale

Digitised Semantics

Representation rich expression of problems and their

solutions structure and logical relations/rules based on generic, re-useable domain

concepts formal, unambiguous, semantics

Rationale for explanation of intent, beliefs and assumptions

Layers of Controlled Natural Languages for human communication

Visualisation for creation and exploration of solutions

Semantic representation for machine processing and formal definition of logical relations

Towards A Solution

Page 10: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[10]

Collaborative Problem Solving Model

basic logic and rationale

Agent, Assumption, ConceptualSpace, Container, Entailment, Inconsistency, PossibleWorld, Proposition, PropositionIndex, Quantity, ReasoningStep, Set, Triple, VarBinding, WorldState

general purpose

ConceptualThing, Constraint, Synchronisation

temporal Precede, TemporalConstraint, TemporalEntity, TimeInterval, TimeLine, TimePoint

space Area, Elevation, Line, Point, SpatialConstraint, SpatialCoordinateSystem, SpatialEntity, SpatialIntersection, SpatialLocation, SpatialUnion

resources Resource, ResourceAllocated, ResourceCapability, ResourceConstraint, ResourceQuantity, ResourceSet

actions Activity, Effect, Precondition

collaborative problem solving

Choice Point, Collaboration, Commitment, Communication, ConstraintViolated, Decision, GoalSpecification, Influence, Issue, JointPersistentGoal, MutualGoal, Problem, Solution, Trust,

planning Allocation, Evaluation, EvaluationCriterion, InitialState, Plan, PlanTask, PlanTaskDescription, PlanTaskTemplate, PlanningProblem, PlanningProblemContext, ResourceCommitment, ResourceReq, TaskCommitment

Copyright IBM UK Ltd, 2009

Page 11: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[11]

Visualisation

Page 12: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

CPM Visualiser

CPM

Solution

Lexicon

Import E

xport

Rule E

xecution

Rules

Graphics

Rationale

Patterns

CPM/OWL

Controlled English

Problem Solving

Concept Modelling

Collaboration

Forms??

Copyright IBM UK Ltd, 2009

Page 13: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[13]

Hybrid Visualisation of RationaleHybrid Visualisation of Rationale

Exploring CNL

Editors

Assumptions, decisions and key facts leading to

resource conflict

Chain of Rationale in CE and Conceptual Graphs

Temporal rationale

Copyright IBM UK Ltd, 2009

Page 14: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[14]

Visualising Rules

A Crate must contain only ammunition of the correct type

“pick any thing from the crate and it have the correct content type”:

if

( the crate C holds the thing S ) and

( the crate C has the value A as content type )

then

( the thing S has the value A as type )

.

Page 15: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[15]

Evolving Design, Evolving Language

15

What do the yellow

areas mean?

How do I represent this in a

language?

the AS90 uses the NATO_L15 for the bombardment at a rate of 2 .

if ( the resource request RR is required by the bombardment T ) and ( the bombardment T has the AS90 A as executor ) and ( the resource request RR requires the NATO_L15 R )

and ( the bombardment T has the value D as duration ) and ( the value Q = the value D * the value 2 )

then ( the resource request RR has the value Q as quantity )

linguistic transformation rule

LOW

HIGH

Copyright IBM UK Ltd, 2009

Page 16: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[16]

Higher Level CNLs

Page 17: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[17]

“Levels of Language”

girl pick fruit. turn. see mammoth. girl run. reach tree. climb. mammoth shake tree. girl yell yell. father run. throw spear. mammoth roar. fall. father take stone. cut meat. give girl. girl eat finish. sleep.

girl pick fruit. turn. see mammoth. she run to tree and climb it. mammoth shake tree. girl yell yell. father run toward her. he throw spear at mammoth. it roar and fall. with stone father cut meat for girl. she eat finish and she sleep.

(from The Unfolding of Language, Guy Deutscher)

elegance, succinctness, specialist

verbose, awkward, genericLOW

HIGH

+ content and function words and grammatical constructs

Page 18: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[18]

at ~~root the ~~noun 1 only ~~verbSing 1 the ~~noun 2 ==> if ( the ~~noun 1 X ~~verbSing 1 the thing Y ) then ( the thing Y is a ~~noun 2 ).

the AS90 only fires the NATO_L15.

if ( the AS90 X fires the thing Y ) then ( the thing Y is a NATO_L15 ) .

LOW

HIGHLinguistic transformation rule

Language transformation

Copyright IBM UK Ltd, 2009

Page 19: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[19]

The “RIF enterprise”

Page 20: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[20]

To express more complex logic of an ontology in a general Mathematical form in an XML syntactic form

To embed the XML syntax in RDF/S/OWL To express the logic in an English like way To represent rationale in the XML syntax

The syntax of the language must have a corresponding formal semantics

All languages in the solution must be formally mappable between each other

Based on standards where possible

Requirements

Page 21: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[21]

Ideally the mathematical logic should be highly expressive, eg First Order Predicate Logic

In practice we may have to accept a less expressive language, especially if based on standards

Therefore Need a full FOPL language as the “gold standard” for

expressing all that we might need Use a subset of FOPL as being the base logic for our

language But must be more expressive than RDF/S/OWL

Requirements - relaxed

Page 22: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[22]

Common Logic

Common Logic as the “gold standard” logic language because:

an ISO standard for FOPL provides a Mathematical form (CLIF) that is readable

(forall … (if (and …) … (earliestfinish t x)))

(ITA) Controlled English for the “human-face” of the logic because:

it is intended as a CNL for CommonLogic we have used it in ITA and have parsers, inferences etc if ( the task T has the value X as earliest start time and has the value MD as minimum duration ) and ( the value X1 = the value X + the value MD ) then ( the task T has the value X1 as earliest completion time ) .

ITA CE

Components selected (1)

Page 23: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[23]

RDF/S/OWL as the semantic web language(s) because it is the SWT standard we have used it already on ITA

RIF as the specification of logic because it is an emerging W3C standard for rule interchange framework for defining different logic subsets focus on definition of semantics Based on:

RIF-FLD RDF compatibility mapping from RIF-BLD to

RDF/S/OWL Other extensions as defined (eg assumptions,

negation)

RIF

RDF/S/OWL

Components selected (2)

Page 24: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[24]

Common Logic

ITA CE

RDF/S/OWL

RIF-FLD

key ITA work to be done

RIF-FLD & CL semantics

RIF-FLD

Sowa

Integration

negation?

Copyright IBM UK Ltd, 2009

Page 25: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[25]

Embedding RIF in OWL

• Don’t want to have different files for RIF and OWL

• Ontology for RIF (“rir”)– rir: Document, rir:Group, rir:Formula, rir:Isa, rir:Frame, rir:And,

rir:Implies, rir:Forall, …

– (rir:Formula used loosely to be Formula or Isa, And, Or etc)

– Variables and Constants interchangeable, via http://rifInRDF#Var and http://rifInRDF#Const datatypes

– Universal rules used to define logical implications, via a fixed Document/Group structure

– Set of universal rules attached to ontology via “LogicSet” entity

Page 26: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[26]

Rationale

Page 27: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[27]

Hmm…What about rationale?

ITA CE

Common Logic

RDF/S/OWL

RIF-FLD

Reasoning Steps,

Assumptions, Decisions,

Facts

Why?

Page 28: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[28]

What is the best standard for representing rationale in all its complexities

(including truth maintenance)

?

Question

Extend RIF?

Bespoke?

PML (from RPI)?

Page 29: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[29]

premise

A implies B A________ B

A implies B

A implies B not B___________ not A

modus ponens modus tollens

Rules of Inference and ReasoningSteps

Logical inference

proposition

A ReasoningStep is an entailment … or an intuition? … or an illogical piece of reasoning?

Entailment

Proposition(logical inference)

Proposition

VarBinding

conclusion

Rule of Inference

Entailment

Reasoning Step

Propositionpremise

Premise propositions

Conclusion propositions

“modus ponens”

Page 30: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[30]

Support and Rationale

• Support is the pathway from propositions (universal, inferred and assumptions) to other propositions via entailments (reasoning steps)

• It is possible to:– generate rationale graphs of a fact, showing the true and false pathways, and the relevant

assumptions and reports.

– Detect incompatible sets of assumptions

– Make and undo assumptions, recalculate the truth values of dependent facts

– Explore possible worlds

E

E

Inferred proposition

True Support

E

False Support

Universal

Assumption

Assumption

entailmentUniversal

Page 31: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[31]

Examples of Rationale

• …but simple concepts lead rapidly to apparently complex support

• Better ways to visualise this are needed ….

• The removal of assumptions can resolve inconsistencies

Page 32: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[32]

Talking about: I assume/decide/believe/because of Solution must have:

explicit support to propositions (ReasoningStep and PropositionIndex)

a “magic” reification step

Proposition (s1 p1 o1)

(s2 p2 o2)

(s3 p3 o3)

Rationale is “talking about”

Proposition

Proposition

because

assume

universal

“Talking about” space RDF Triple space

reificationsupport

RS

PI

PI

Page 33: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[33]

Approaches to Proposition reification

• Original CPM:

• Freeform:

• CE:

• RIF:

• RDF graph:

P

PPAND

Triple

Triple

(s1 p1 o1)

(s2 p2 o2)

P “I decided to move the tanks forward over the hill”

P “the task T realises the objective O”

P Formula … FrameSlot (s1 p1 o1)

(s1 p1 o1)CL

RIF

informal

P “<rdf:rdf> …. </rdf>” (s1 p1 o1)RDF(s2 p2 o2)

(s2 p2 o2)

Copyright IBM UK Ltd, 2009

Page 34: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[34]

Truth Values (New)

• Truth is not defined by the existence of an RDF triple but the rationale support

• ATMS “label” permits efficient calculation of truth conditions of a proposition (hence triples)– Proposition P: (OR (AND A1 A2) A3 (A5 A6)) where AN are

(atomic?) propositions

• RIF: the label as a property of Proposition:

• Bespoke efficient property of Proposition:

• CE sentence: that <P> is supported by the assumption that <A1> and that <A2> or by the assumption that <A3> or by the assumption that <A5> and that <A6>

P Formula OR AND ANsupport

P support “URI_A1,URI_A2|URI_A3|URI_A5,URI_A6”

Page 35: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[35]

Summary of rationale using RIF

• “RIR”: a set of RDF types and properties to represent a RIF-FLD ontology embedded in an RDF document

• An Entailment that has:– Premise propositions (including rules)

– Conclusion propositions

– Variable bindings

• A Proposition has truth support defined by PropositionIndex(es)– permits defeasible reasoning and paraconsistency, possible

worlds, modelling of agents beliefs, etc

• A Proposition points to rir:Formula (isa, frameslot…)

• The rir:Frameslot, rir:Isa (#), rir:Subclass (##) reify to RDF triples, as per the RIF compatibility document

Page 36: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[36]

Rationale

Isa

Common Logic

ITA CE

RDF/S/OWL

RIF-FLD

ReasoningStep

Entailment

RuleOfInference

PropositionIndex

(Rule) Proposition

Frameslot

(s p o)

reifiestype of

rule of inference

Rationale using RIF

Proposition

truth conditionsEMBED

agent

Assumption

VarBinding

premises/conclusions/

bindings

Subclasssupport

Formula

Copyright IBM UK Ltd, 2009

Page 37: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[37]

RATIONALE

Integration of Rationale

Common Logic

ITA CE

RDF/S/OWL

RIF-FLD

key ITA work to be done

RIF-FLD & CL semantics

RIF-FLD

Sowa

Copyright IBM UK Ltd, 2009

Page 38: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[38]

Negation

Page 39: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[39]

(Tentative) Approach to Negation

• Create RIF Dialect– Retain RIF-FLD mapping to RDF/S/OWL

• Symmetric negation as classical negation:– Neg operator

– CE: “it is false that …”

– Neg premise only matches a Neg fact (?)

– Classical semantics

• Default negation as assumption based default reasoning– Naf operator

– CE: “if condition1 and it is assumed false that … then …

– Assumption permitted as long as its not inconsistent (or use ATMS)

– Etherington’s semantics for default logic

Page 40: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[40]

Semantic Web

Page 41: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[41]

Semantic & User Interface Research

• How can the semantic web be used to:– harness collective intelligence

– support the collective endeavour of groups of people

• Semantic Wikis– Use of Semantic MediaWiki and a CNL interface to

collaboratively construct ontologies

• Graphical Queries– graphical drawing of SPARQL queries

– Visual Query Builder at the level of the conceptual model

• Semantic Web techniques– SWEDER - Semantic Wrapping of data sources and rules

– GIDS – Framework for distributed access to interlinked data

Page 42: International Technology Alliance In Network & Information Sciences International Technology Alliance In Network & Information Sciences Dr David Mott IBM

[42]

Your Work

How can we help the community?

How can we collaborate?