Distributed Imprecise Design Knowledgeon the Semantic Web

Julian R. Eichhoff 1 and Univ.-Prof. Dr.-Ing. Wolfgang Maass 2

1 Furtwangen University, Furtwangen, Germanyjulian.eichhoff@hs-furtwangen.de

2 Saarland University, Saarbrücken, Germanywolfgang.maass@iss.uni-saarland.de

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 2

Motivation:Support Lead-Qualification for Solution Selling

Solution Selling: offer customized products and services, which meet customer's individual business needs

Constraints

Goals

Solution Implementation

Services

Goods

DeploymentCustomization& integration

Requirements definition

Postdeployment support

Lead Qualification

ImplementedSolution

cf. Tuli, K.R., Kohli, A.K., Bharadwaj, S.G.: Rethinking Customer Solutions: From Product Bundles to Relational Processes. J. Market. 71(3), pp. 1-17 (2007)

Salesperson Projection:Which good/service bundles may serve as a solution?

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 3

Proj

ect M

gmt.

Use

r Gro

upSo

ftw

. Eng

.U

ser G

roup

FBS Bayesian Network (FBS-BN)

Floo

r 10

Offi

ce

Function Concepts Behavior Concepts Structure Concepts

DESIGN OBJECT ATTRIBUTES

Adjustability of Floor 10 Layout of Floor 10

Being Flexiblewrt.

Floor 10

DES

IGN

OBJ

ECT

COM

PON

ENTS

yes

no

0.5

0.5

Is Being Flexible important for your

customer?

yes no

low 0.5

high 0.5

low 0.0

high 1.0

Is Being Flexible important for your

customer?

yes no

low yes 0.0

high yes 1.0

low no 0.5

high no 0.5

low yes 0.0

high yes 1.0

low no 0.5

high no 0.5

yes 1.0

no 0.0

yes 0.0

no 1.0Answer concepts

implicitly

Search for uncertain concepts

cf. Eichhoff, J.R., Maass, W.: Representation and Reuse of Design Knowledge: An Application for Sales Call Support. In: Konig, A., Dengel, A., Hinkelmann, K., � Kise, K., Howlett, R.J., Jain, L.C. (eds.) KES 2011, Part I. LNAI, vol. 6881, pp. 387–396. Springer, Heidelberg (2005)

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 4

FBS Bayesian Network (FBS-BN)

Floo

r 10

Offi

cePr

ojec

t Mgm

t.U

ser G

roup

Soft

w. E

ng.

Use

r Gro

up

Function Concepts Behavior Concepts Structure Concepts

DESIGN OBJECT ATTRIBUTES

Adjustability of Floor 10 Layout of Floor 10

Being Flexiblewrt.

Floor 10

DES

IGN

OBJ

ECT

COM

PON

ENTS

mismatch low low 0.0mismatch low high 0.0mismatch high low 0.0mismatch high high 0.0

match low low 1.0match low high 0.0match high low 0.0match high high 1.0

mismatch low low 0.0mismatch low high 1.0mismatch high low 1.0mismatch high high 0.0

match low low 1.0match low high 0.0match high low 0.0match high high 1.0

Assess problem/solution space

Evaluate differentsolutions

Evidence Propagation

Match of and ? ( )

Evidence Propagation

Evidence Propagation

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 5

FBS Bayesian Network (FBS-BN)

Floo

r 10

Offi

cePr

ojec

t Mgm

t.U

ser G

roup

Soft

w. E

ng.

Use

r Gro

up

Function Concepts Behavior Concepts Structure Concepts

DESIGN OBJECT ATTRIBUTES

Enclosure of Floor 10 Partitions on Floor 10

Adjustability of Floor 10 Layout of Floor 10

Being Flexiblewrt.

Floor 10

Distractions of Project Mgmt.

Quiet Work of Project Mgmt.

Being Efficientwrt.

Project Mgmt.

Distractions of Softw. Eng.

Quiet Work of Softw. Eng.

Being Efficientwrt.

Softw. Eng.

DES

IGN

OBJ

ECT

COM

PON

ENTS

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 6

Distributed Use of FBS-BN-Representations

Clie

ntSh

ared

RD

F-St

ore

Design Situation Design Knowledge

instantiates

templateForFBS-BN

Knowledge Engineering App.

canBeRelated

ComponentClasses

connect

FBS-Relations

FBS-Concepts

reifi

cate

desc

ribe

connect

isRelated

ComponentIndividuals

connect

inst

ance

sOf

I/O

CPDs

Variables

reifi

cate

desc

ribe

connect

I/O

Knowledge Reuse App.

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 7

Distributed Use of FBS-BN-Representations

canBeRelatedTo

attributeOf

ComponentClass

instanceOf

isRelatedTo

hasSelection

ComponentIndividual

reflectsInfluenceOf/On

expressesBeliefIn

parameterOf

hasValue

Distribution

Proability

Value

expe

cts

implicates

exhi

bits

implicatesimplicates

Behavior StructureFunction

describedBy categorizedByDescription State

definedAs

Conceptualization

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 8

Future Work

• Evaluate prototypes empirically• Test large scale network inference• Apply model in different sales settings• Integrate machine-learning• Conceptualize and implement module for

explanation of inference results

Julian R. Eichhoff & Univ.-Prof. Dr.-Ing. Wolfgang Maass14.09.2011 Slide 9

julian.eichhoff@hs-furtwangen.de

http://iss.uni-saarland.de

Questions?