PROVISION OF CONTEXT-SENSITIVE ENTERPRISE

KNOWLEDGE FOR DECISION SUPPORT

An Approach based on Enterprise Models and Information Demand Contexts

Tatiana Levashova, Michael Pashkin

St.Petersburg Institute for Informatics and Automation, 39, 14th line, St.Petersburg, 199178, Russia

Magnus Lundqvist

School of Engineering at Jönköping University, Gjuterigatan 5, SE-551 11 Jönköping, Sweden

Keywords: Ontology, constraint networks, context sensitive decision support, enterprise model, information demand.

Abstract: In this paper an approach for deriving abstract and operational context for context-sensitive decision

support, and thereby also parts of information demand contexts, from enterprise models is presented

together with some thoughts on how this can be utilised in the efforts of trying to provide users with current,

correct, and relevant information with respect to the tasks such users perform within organisations. The

different steps involved in the process of deriving context from enterprise models is explained by means of

different representations of an example model produced in earlier research done by the authors.

1 INTRODUCTION

The idea of supporting business professionals or

“white-collar workers” in the solving and

performance of problems and work related tasks by

means of different types of information systems is

well established today. However, despite the ever-

increasing number of systems and approaches tried

many of the old issues related to getting the “right”

information for specific tasks or problems to the

“right” user remain. On top of this many of the

systems and approaches intended to improve the

situation have introduced a number of new problems

and thus to some extent made the situation worse.

Today the problem no longer is to find information

but rather how to select the relevant information

from the massive amount of information available in

and to organisations. As it has been shown in several

investigations done over the years, one of the bigger

ones being Delphi Group’s (2002), the current

situation regarding information (over)flow, reuse,

and organisational memory is problematic at best.

1.1 Related Research

The research presented in this paper builds heavily

on work within the areas of Information Logistics,

Information Demand modelling and analysis, and

Decision Support as well as the combination of these

areas.

Several similar approaches to demand driven

information supply for supporting decision-making

exist. One example is a methodology for information

requirements analysis for data warehousing systems

(Winter and Strauch, 2003). The methodology

supports the entire process of determining

information requirements for a targeted decision

process. The information requirements are defined

as the type, amount, and quality of information that a

decision maker needs to do his/her job. The

methodology starts with an analysis of actual

information supply and creation of an information

map. Information demand is determined based on

typical questions relevant to the targeted decision

process. The information demand derived based on

these questions is matched against the information

map. The matching reveals non-covered information

requirements.

The DDIU (Data Demand and Information Use)

conceptual framework (Foreit et al., 2006) does,

much in the same manner, supports evidence-based

decision making for health systems. The DDIU

framework explains the context in which decisions

are made and how this context influences the

demand for data, the use of information, and the

Levashova T., Pashkin M. and Lundqvist M. (2007).

PROVISION OF CONTEXT-SENSITIVE ENTERPRISE KNOWLEDGE FOR DECISION SUPPORT - An Approach based on Enterpr ise Models and

Information Demand Contexts.

In Proceedings of the Ninth International Conference on Enterprise Information Systems - AIDSS, pages 88-93

DOI: 10.5220/0002396700880093

 SciTePress

collection and availability of data. Information

demand in this framework is specified by decision

makers or other stakeholders. Data resources are

aligned with the decisions they would support, and

vice versa. For identification of gaps (areas where

useful data were readily available but not used) in

the information flow a visual context is provided.

The research concerning information demand

issues and the research being done by the authors

individually have a lot of core concepts in common.

As a consequence, it was decided to put effort on

trying to combine the individual approaches of the

authors in order to produce added value to both areas

in terms of more efficient analysis methods, tools

and systems.

2 INFORMATION DEMAND

As an attempt to solve at least some of the problems

mentioned in the introduction, the area of

Information Logistics has been suggested as a way

to “strike” at the core of information overflow

related problems. Achieving this is attempted by

making sure that, users of information systems are

provided only with the “right” information at the

“right” time and place (Deiters et. al., 2003).

However, right information at the right time and

place implies that something must be known about

the users and the needs motivating their use of an

information system. That is, right information, time,

and location require some knowledge of what right

means for the specific user with respect to his/her

current situation and problem(s).

As a consequence of these implications different

methods for the analysis and modelling of informa-

tion demand (Lundqvist, 2005) as well as a number

of related definitions (Lundqvist and Sandkuhl,

2004), has been developed by the Information Engi-

neering Group at Jönköping University.

The core concept of Information Demand

Analysis aims at deriving what is referred to as

Information Demand Context (IDC), capturing the

context in which information demands exist, i.e.

deciding what the –right– in relation to information,

time, and location refers to. It has been argued that

such IDCs can be derived from a number of sources

within an organisation but it has also been suggested

that a particular suitable source for this is Enterprise

Models (EM) (Lundqvist, 2005) describing many or

all aspects of an organisation, e.g. products,

processes, organisational structures, resources etc.

Such an approach to deriving IDC has, as has been

shown in earlier work made by the authors

(Levashova et. al., 2005), many similarities to the

area of context-sensitive decision support. In the

joint research a model of a fictitious bike producing

enterprise has been used as an exemplifying EM.

3 CONTEXT-SENSITIVE

DECISION SUPPORT

The purpose of the presented approach to decision

support is to model a situation or a problem

formulated by a user as requests to a Decision

Support System (DSS). The problem or the situation

is modelled in terms of context. Two types of

context are used; abstract and operational. Abstract

context is an ontology-driven model integrating

information and knowledge relevant to the solving

of the problem. Operational context is an

instantiation of the abstract context.

DSS according to this approach has two distinct

main phases: (1) the preparatory phase and (2) the

decision making phase. At the preparatory phase

models of DSS components (domain ontology, user

profile, and information sources) are created. All the

models are represented by means of an object-

oriented constraint networks (OOCN) formalism.

The domain ontology is supplied with links to the

user profiles and information sources. The links

mean that class attributes of the ontology get values

provided by the users or information sources. At the

decision making phase abstract and operational

contexts are produced.

According to the OOCN formalism, an ontology

is represented by a set of classes; a set of class at-

tributes; a set of attribute domains; and a set of con-

straints. The set of constraints comprises

(1) taxonomical (“is-a”) relationships,

(2) hierarchical (“part-of”) relationships, (3) class

cardinality restriction, (4) class compatibilities,

(5) associative relationships, and (6) functional rela-

tions. Support for this formalism is included in the

Web-DESO tool (Smirnov et al., 2002).

The tool used for modelling the bike producing

enterprise uses XML as an internal representation of

information that is compatible with the OOCN

formalism. Table 1 shows correspondences between

the EM, its representation in XML, and the OOCN

formalism.

In the first attempts of combining the different

approaches to Information Demand Analysis and

Context-sensitive Decision Support a file containing

such an XML-representation of EM was used as the

source from which the domain ontology (Figure 1)

was derived (Levashova et al., 2006b). It comprises

around 500 classes, 2500 class attributes, 400 part-of

relationships, and 850 associative relationships. The

domain ontology does not contain instances though.

PROVISION OF CONTEXT-SENSITIVE ENTERPRISE KNOWLEDGE FOR DECISION SUPPORT - An Approach

based on Enterprise Models and Information Demand Contexts

Business

Operation

Organization

Processes

Organization

Role

Purchaser

Assembler

Distributor

Business

Processes

Supply chain

Bike

production

Purchasing

Assembly Distribution

Part-of

Is-a

Root

High-level

processes

Production

planner

Business

Functions

Product Development

and Manufacturing

Corporate Business

Administration

Marketing and Sales

Supply

Distribution

Production

planning

A hierarchically organized Organization with a number of formal positions and a number of roles that can be performed by individuals

or Organization Units, as well as some key people for the kind development that the model is covering.

Organization units that constitute the business and some of its high-level positions. Positions are formal elements of an Organization.

IT Resources,

Standards and Products

IT Support Personnel

Person

Geography

and facilities

Location

Latin America

Europe

…

Figure 1: Domain ontology (part of): taxonomy and hierarchy views.

Table 1: Correspondence between Enterprise Model,

XML, and OOCN representations.

Attribute values can be calculated by the DSS based

on the set of functional constraints specified in the

ontology or they can be taken from information

sources. The EM contains references to various files

that provide the required information.

For example, in the domain ontology part

(Figure 1) the class “Geography and facilities”

defines the locations that comprise the enterprise.

The use of the object type ranges from individual

rooms, buildings, and cities to countries and regions.

These locations are stored in a database that is

considered to be an information source. This

information source provides values to the attributes

specified in the class “Location” and its subclasses.

Access to the information sources is provided

through Web-services, which are responsible for the

interaction with these sources in the following ways:

• representation of information sources by means

of the OOCN formalism;

• querying information sources;

• transfer of the information to the DSS;

• information integration;

• data conversion.

Figure 2 presents some screens of the

Web-DESO tool illustrating the way of linking the

domain ontology and information sources in general

and the database with locations in particular. In the

domain ontology the method for the location

definition is introduced (lower screen). This method

connects the Web-service responsible for the

interactions with the database of locations by

passing the URI of the Web-service as an argument

to the method. The OOCN-formalism then handles

method arguments as attributes. The method outputs

the name of the location and its address. The name

of the location is the name of a plant, an office, or

some other organisation located at a certain address.

This name along with the other address

characteristics will instantiate the attributes specified

in the class “Geography and facilities” (upper

screen). This is specified through functional

constraints. In earlier joint research efforts

(Levashova et al., 2006a) abstract contexts

describing the role activities within the enterprise

EM XML OOCN

Object Object Class

Container Object Class

Instance Value set A set of

attributes

Range Data type Domain

Type-of;

Child-link

“Is-a”

constraint

Parent-children

relationship

Part-link “Part-of”

constraint

Named relationships but

“Type-of”/“Of-type”

Relationship Associative

constraint

ICEIS 2007 - International Conference on Enterprise Information Systems

Figure 2: Linking domain ontology and an information source.

were produced for the roles introduced in EM

(Figure 1, class “Role”). Those were rather large

contexts describing what information one filling a

certain role needs on the whole. Moreover, since

solving a specific task as part of a certain role often

involve and affect other roles, and tasks related to

them, the contexts included practically all the

information the EM contains.

For the purpose of this paper an exemplifying

task, production planning (the shaded box in

Figure 1) performed by the role production planner,

has been chosen as an illustration of how the ideas

presented here can be used in practice. A context

modeling this task will include knowledge relevant

only to this particular task within the specific role.

Abstract context is produced from the domain

ontology based on a set of keywords corresponding

to the meaningful words contained in the user’s

request to DSS. Knowledge surrounding the

keywords is captured so that it would be relevant to

the user request and to the inference supported by

OOCN. Here producing abstract context is

illustrated for the keywords “production planner”

and “production planning”. “Production planner” is

the name of a role, “production planning” is a name

of an activity, among others, performed by this role.

Unlike the domain ontology, the main purpose of

the abstract context is to collect knowledge relevant

to the task rather than to provide a well-formed

knowledge classification. The abstract context

produced for the task in question is a two-level, class

taxonomy (Table 2) expanded with part-of,

associative, and functional relationships. The classes

of the top level are used as a way to determine the

type of knowledge the classes of the bottom level

belong to. The top-level classes specify inherited

attributes for the classes of lower levels. The classes

of the bottom level are the ontology classes most

specified; they provide a complete specification for

the instances of these classes.

An explanation of the knowledge included in the

abstract context is given below. The explanation

aims at clarifying the associations between the

classes that are not given in Table 2. According to

the OOCN formalism the associations in the context

are represented by associative relationships. The

words corresponding to this kind of relationships are

italicised in the text that follows.

Production planning is a bottom line process for

the business process of bike production. The bike

production process is used by the business function

of product development and manufacturing. The

input for the bike production process is information

about the sales plan. The bike production is affected

by a business strategy for the improvement and

expansion of research and development. In

production planning application functions and

services of commodities and product purchasing are

used. These services are provided by the

applications listed in Table 2 in the “Bottom level

class” column for the “Application” row. All of

them are of the mainframe type. The applications

PROVISION OF CONTEXT-SENSITIVE ENTERPRISE KNOWLEDGE FOR DECISION SUPPORT - An Approach

based on Enterprise Models and Information Demand Contexts

Table 2: Classes entered into the abstract context for the task of production planning.

are located in different countries in several plants

and offices that are instances of the geography and

facilities class. Some of the applications use a

specially developed application building block,

providing some system modifications.

The applications need IBM mainframe-based

database system for relational databases “DB2” and

COBOL language. The applications are

implemented by IBM MRP (mainframe-based

product for production planning) and the

applications are supported by the members of the

IT-personnel that are instances of the class “Person”.

In terms of information demand the abstract

context produced can be interpreted as follows: the

production planner needs, for performing the task of

production planning, a sales plan, information about

relevant parts of the business strategy, applications

(and their building blocks) supporting planning

functions, certain databases and programming

languages, and to some extent also personnel with

competence and skills that might contribute to the

planning activities.

Producing operational context, i.e. providing the

user with the needed information, is then done by

selecting attribute values from the information

sources and computation of the related values. For

instance, in order to obtain a sales plan the produc-

tion planner has to indicate the planning period that

he/she is interested in. Referring to the example with

application locations the production planner can

assign some criteria to select the location of the ap-

plications he/she intends to use. This selection (and

only it) influences the value for the attribute of “the

current cost of IT support for the process” specified

in the class “Production planning”.

If the user profile contains some statistics based

on which user preferences that have been identified

the selection criteria can be taken from this profile.

In other cases they have to be entered directly by the

user. For this a special user interface is used, one

that is developed specifically for each case but since

the enterprise model used as a basis for the research

done so far describes a fictitious enterprise no such

interface exists yet. Based on the user criteria, DSS

sends requests to the Web-services.

Top level class Description Bottom level class

Bike production A collection of processes to purchase parts,

assemble and distribute bikes

Production Planning

Business Functions Functions of the business, normally used at a very

high level

Product development and

manufacturing

Information Flow Information transferred from one process to another Sales plan

Strategies A statement controlling what the business intends to

do to achieve its goals and objectives

Improve and expand research and

development

Commodities purchasing Application Function A function or service provided by one or more

applications to assist one or more user processes

Product purchasing

BUYBUY - Purchasing

Resource consumption reporting

Operations preparation and planning

Warehouse management

Mtrl Req'ts

Application The business applications, operational, planned, and

phased out

SALFOR - History-based sales

forecaster

Geography and

facilities

Geography defines the locations that comprise the

enterprise. The object type ranges from individual

rooms, buildings, and cities to countries and regions

Location

Our purchasing system modifications Application Building

Block

An architectural element used to build one or more

applications

Our MRP modifications

Mainframe A mainframe-based computer, consisting of a

hardware platform and an operating system

Mainframe

IT Support Personnel The personnel available for IT maintenance and

development

Person

DB2 Competence/Skill An object describing the condition and areas of

human capabilities or abilities

COBOL

Application Product A named piece of business-oriented software offered

by an external or an internal vendor

IBM MRP

ICEIS 2007 - International Conference on Enterprise Information Systems

Finally the Web-services query the information

sources, get query results, and then pass them on to

the operational context.

4 CONCLUSION

During previous performed research referenced in

this paper it has become clear that Information

Demand Analysis and Modelling as an area (and

thereby also Information Logistics) has a lot to

benefit from applying ideas, concepts, and

techniques from the context-driven approach to

decision support. However, research performed by

the authors has up to this point focused more on

deriving information demand contexts and thereby

identifying information demands rather than on

matching information demands and information

present in various kinds of enterprise systems. In this

paper some first ideas concerning how the matching

between information and information demands can

be performed within decision support systems based

on the initial generation of IDCs from an enterprise

model has been presented.

It is believed that continued research on this

subject could prove to be most useful for informa-

tion logistic systems as well as for decision support

systems since it could lead not only to the automatic

identification and derivation of information demands

in relation to work tasks but also the matching of

such demands to enterprise information as well as

semi-automatic provision of such information, all

this based on one single enterprise model. If suc-

cessful such an approach would not only greatly

decrease the amount of analysis required when

building systems supporting knowledge and infor-

mation intensive work-related tasks, it could also

contribute to the reduction of information overflow

in everyday working situations.

ACKNOWLEDGEMENTS

Parts of the presented research are due to CRDF

partner project # RUM2-1554-ST-05 with US ONR

and US AFRL, projects supported by the Russian

Academy of Sciences # 16.2.35 and # 1.9, grant

# 05-01-00151 of the Russian Foundation for Basic

Research, and grant # IG 2003-2040 from the

Swedish Foundation for International Cooperation in

Research and Higher Education STINT.

REFERENCES

Deiters, W., Löffeler, T., Pfenningschmidt, S., 2003. The

Information Logistical Approach Toward a User

Demand-driven Information Supply. In Spinellis, D.

(Ed.). Cross-Media Service Delivery. Kluwer

Academic Publisher, 47 – 38.

Delphi Group, 2002. Perspectives on Information

Retrieval. Retrieved November 20, 2006 from Delphi

Group Web site: http://www.delphigroup.com/-

research/ir_perspectives_sum.pdf.

Foreit, K, Moreland, S, LaFond, A., 2006. A Conceptual

Framework for Data Demand and Information Use in

the Health Sector. Retrieved March 7, 2007, from the

University of North Carolina, MEASURE Evaluation

Web site: https://www.cpc.unc.edu/measure/-

leadership/ddiu.html.

Levashova, T., Lundqvist, M., Sandkuhl, K., Smirnov, A.,

2006a. Context-based Modelling of Information

Demand: Approaches from Information Logistics and

Decision Support. Proceedings of the 14

European

Conference on Information Systems. Gothenburg,

Sweden, CD-ROM.

Levashova, T., Lundqvist, M., Pashkin, M., 2006b.

Moving towards Automatic Generation of Information

Demand Contexts: an Approach Based on Enterprise

Models and Ontology Slicing. Proceedings of OTM

Confederated Intl. Conferences CoopIS, DOA, GADA,

and ODBASE 2006, Montpellier, France, 1012 – 1019.

Lundqvist, M. & Sandkuhl, K., 2004. Modelling

Information Demand for Collaborative Engineering.

Proceedings of 2

Intl. Workshop on Challenges in

Collaborative Engineering. Stara Lesna, Slovakia:

VEDA, 111 – 120.

Lundqvist, M. 2005. Context as a Key Concept in

Information Demand Analysis. Proceedings of the

Doctoral Consortium associated with the 5

Intl. and

Interdisciplinary Conference on Modelling and Using

Context (Context-05), Paris, France, 63 – 73.

Lundqvist, M., Sandkuhl, K., Levashova, T., Smirnov, A.,

2005. Context-Driven Information Demand Analysis

in Information Logistics. Proceedings of the first

International Workshop on Contexts and Ontologies:

Theory, Practice and Applications. Pittsburgh,

Pennsylvania, USA, AAAI Press, 124 – 127.

Smirnov, A., Pashkin, M., Chilov, N., Levashova, T.,

2005. Operational Decision Support: Context-Based

Approach and Technological Framework. Proceedings

of the 5

International Conference CONTEXT-05,

Paris, France, LNAI 3554, Springer. 476 – 489.

Smirnov, A., Pashkin, M., Chilov, N., Levashova, T.,

2002. "Web-DESO": Ontology Management System.

Intl. Conference on Intelligent Information

Technology (ICIIT’2002), Beijing, China, 485 – 494.

Winter, R., Strauch, B., 2003. Demand-driven Information

Requirements Analysis in Data Warehousing. Journal

of Data Warehousing, 8, 1, 38 – 47.

PROVISION OF CONTEXT-SENSITIVE ENTERPRISE KNOWLEDGE FOR DECISION SUPPORT - An Approach

based on Enterprise Models and Information Demand Contexts