LOCATING KNOWLEDGE THROUGH AUTOMATED

ORGANIZATIONAL CARTOGRAPHY [AUTOCART]

Mounir Kehal, Sandrine Crener, Patrice Sargenti

CSITDS Research Group

International University of Monaco

Monte Carlo, Principality of Monaco

Keywords: Knowledge Management, Knowledge Maps, Neural Networks, Organizational Cartography, Semantic

Relevance, Unsupervised learning, Kohonen Networks, Self-Organizing Maps (SOMs).

Abstract: The Post-Globalization aeon has placed businesses everywhere in new and different competitive situations

where knowledgeable, effective and efficient behaviour has come to provide the competitive and

comparative edge. Enterprises have turned to explicit- and even conceptualising on tacit- Knowledge

Management to elaborate a systematic approach to develop and sustain the Intellectual Capital needed to

succeed. To be able to do that, you have to be able to visualize your organization as consisting of nothing

but knowledge and knowledge flows, whilst being presented in a graphical and visual framework, referred

to as automated organizational cartography. Hence, creating the ability of further actively classifying

existing organizational content evolving from and within data feeds, in an algorithmic manner, hence

potentially giving insightful schemes and dynamics by which organizational know-how is visualised. It is

discussed and elaborated on most recent and applicable definitions and classifications of knowledge

management, representing a wide range of views from mechanistic (systematic, data driven) to a more

socially (psychologically, cognitive/metadata driven) orientated. More elaborate continuum models, for

knowledge acquisition and reasoning purposes, are being used for effectively representing the domain of

information that an end user may contain in their decision making process for utilization of available

organizational intellectual resources.

1 INTRODUCTION

Approaches to manage knowledge have been largely

based on various combinations of business practices,

management strategies, and subject related research.

Examples of these approaches are innumerable and

include organizational learning, the learning

organization, total quality management (TQM),

business process re-engineering (BPR), quality

circles (QCs), and so on. Of more recent times,

especially in the last decade or so, Knowledge

Management (KM) has started to emerge as

multidisciplinary area of interest in academia and

business worlds. We cover and provide a framework

of how knowledge may be modelled; thus specified,

for the development of information systems

supporting attempts to manage knowledge.

2 KNOWLEDGE MANAGEMENT

LITERATURE SYNOPSIS

While definitions of any subject matter can be

helpful in regard to clarifying the scope and depth of

the subject under consideration, they can also be

notoriously difficult to articulate. Some authors in

the field have tried to provide a significant and

diverse range of definitions for knowledge.

Hedlund, for example, used ‘knowledge’ and

‘information’ interchangeably and although he

acknowledged that they should be distinguished, his

use amounts to treating them as identical (Hedlund,

1994). Nonaka and his colleagues describe

knowledge as ‘a meaningful set of information that

constitutes a justified true belief and/or an embodied

technical skill (Nonaka et al, 1996). We may

consider Knowledge Management as a framework

providing the ability to utilize the available

knowledge resources effectively, and in a timely

351

Kehal M., Crener S. and Sargenti P. (2006).

LOCATING KNOWLEDGE THROUGH AUTOMATED ORGANIZATIONAL CARTOGRAPHY [AUTOCART].

In Proceedings of the Eighth International Conference on Enterprise Information Systems - AIDSS, pages 351-355

DOI: 10.5220/0002467903510355

 SciTePress

manner, for organizational benefit and advantage.

Essentially, it may be evident in organizational

processes, the combination of data and information

sources, the processing capacity of IT solutions,

people, and the creation and innovative sharing of

knowledge throughout the organization. Such

framework would inevitably lead to a true managing

of knowledge, on a contextual basis that maximizes

the utilization behind available know-how, -why, -

what, -when, -where, -who.

2.1 Knowledge Category Models

Such types of model categorize knowledge into

discrete elements. For instance, Nonaka’s model is

an attempt at giving a high level conceptual

representation of KM and essentially considers KM

as knowledge creation process. Figure 1 shows

Nonaka’s knowledge management model reflecting

knowledge conversion and dissemination modes.

Tacit Explicit

Tacit

From

Explicit

Figure 1: Nonaka and Takeuchi’s Knowledge

Management model (Nonaka et al, 1995).

As can be observed from the figure above,

knowledge would be composed of two constituents,

Tacit and Explicit. Tacit Knowledge is defined as

non-verbalized, intuitive, and unarticulated. Explicit

or articulated knowledge is specified as being

formally structured in writing or some pre-defined

form. Nonaka’s model assumes tacit knowledge can

be transferred through a process of socialization into

tacit knowledge and that tacit knowledge can

become explicit knowledge through a process of

externalisation. The model also assumes that explicit

knowledge can be transferred into tacit knowledge

through a process of internalisation, and that explicit

knowledge can be transferred to explicit knowledge

through a process of combination. In relation to the

knowledge conversion model transcribed in Figure

1, we believe that knowledge creation undergoes a

nested set of computerized processes [explicit] and

accompanying practices [tacit], allowing as well for

its interlinkages and cross levelling to diverse

specialist areas of expertise and to those it would

tend to restrain, as knowledge would be considered

as highest level available for awareness on the

object of concern. Hence, aim is rather to acquire

automatically, represent visually, and reason

collectively on textual content contained. Thus, a

computationally mediated tool is conceptualised

upon subsequently, being referred to as

AUTOCART, AUTomated Organizational

CARTography, supporting knowledge evolution

studies, knowledge sharing and corresponding flow

representation.

3 ORGANIZATIONAL

CARTOGRAPHY AND

KNOWLEDGE MAPPING

According to Oxford English Dictionary,

Cartography is the drawing of charts or maps. Our

aim is to generate cartograms representing stored

content attained from specialist data feeds. Figure 4

represents, the characteristics by which ‘information

in context’, knowledge, is dealt in the process of its

acquisition. From internal to external sources, and

from being data that is interpreted, to one that

models certainty with intent to validate its semantics

by knowledge workers.

Certainty

Lo Med Hi

Hi Hi

Internal Med Med External

Lo Lo

Lo Med Hi

Interpretation

Figure 4: Knowledge Acquisition Spectrum

Hence from Figure 4, Certainty, Internal,

Interpretation and External are all knowledge

instances attained by means of capturing tacit and

explicit knowledge, with possibly varying values,

states and roles, from knowledge workers, and the

levels of processing achieved by a mediated

computation. Figure 5, below reflects the nature

anticipated by such processing in a framework that

models parameters of consideration from which

knowledge may be viewed, or rather represents and

Socialization Externalisation

Internalisation Combination

ICEIS 2006 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

352

embeds itself in the form of an [intangible object]

action, thinking, [tangible object] archetype, human.

Action

Lo Med Hi

Hi Hi

Archetype Med Med Human

Lo Lo

Lo Med Hi

Thinking

Figure 5: Knowledge Conversion Spectrum

4 SYSTEMATIC VIEW OF

AUTOCART

The knowledge spectrum models covered above

would provide us with a framework for the

development of AUTOCART, represented at a finer

degree of abstraction in Figure 6 AUTOCART

Meta level model, by use of dependency

relationships and associations among processes

and/or instances of objects. The Relationships and

associations are stereotyped as <<refine>>, in

accordance with the UML (Unified Modelling

Language) notation (Booch et al, 1999).

Figure 6: AUTOCART, Meta Level model.

These dispositions of knowledge comprise parts of

the Knowledge Constituents, which embody the

‘raw’ material of the organisation in question.

Therefore, a generalisation relationship is used to

depict the more specific kinds of knowledge

elements in relation to the ‘whole’. Knowledge

Constituents undergo some form of filtering, based

on criteria derived from the document specification

model and partly determined by the textual content.

These functional processes are modelled in the next

model, Figure 7, which focuses on functional

requirements at a lower level of processing. In like

manner, each knowledge element, texts in this case,

is assigned its textual category, primarily

determined by its textual contents opted for in a pre

defined algorithmic manner, using principles of

Kohonen Nets, for instance; and directions towards

an automated learning environment through

Certainty

Explicit and

Tacit

Internal

External

Inter

retation

Filter

/ Textual content

Knowledge

Constituents

Text Categories

/ Document specification

ode

Knowledge

Nodes

Human Innovation

Archetype Action

undergo

assign to

form

<<Refine>>

/ or generate

LOCATING KNOWLEDGE THROUGH AUTOMATED ORGANIZATIONAL CARTOGRAPHY [AUTOCART]

353

induction and hence possible alterations in terms of

activation and threshold functions deterministic

weights, leading to toggling between unsupervised

and system-supervised learning for a networked

representation of data. To establish the textual

category is vital in classifying textual content and,

among with characteristics such as links, directly

added from the filtering process, forming the basis

of a knowledge node, being interlinked using a

generalisation relationship, following the notation of

UML (Booch et al, 1999).

Figure 7: AUTOCART, process level.

At process level, Data Elements within Knowledge

Constituents are to be filtered and then accordingly

classified. The overall aim is to surface the latent

semantic structure of the Knowledge Constituents.

The filtering process is primarily based on a

document specification model – incorporated in

Figure 6 – which is an aggregate of textual

components. These can be identified as being the

actual text of the document, annotation apparent in

the document and the links present. The latter can

be further specialised into association Links –

pointing to and from related documents – and

classification Links, including domain, project and

user specific links, and other relationship links as a

build-up of the data semantics is incurred based on

semantics of content. Once text components have

been determined, each text is assigned a Text

Category, driven by the cohesive relation between

the document specification model and the textual

content. In case the category is not readily known, a

Category Generator is invoked, whereby assigns a

category in an algorithmic manner. Effectively, the

process of textual categorisation and filtering results

in assigning some kind of index to each textual input

– in the form of data entries per document – in an

attempt to reveal the latent semantic structure

underlying the organisational knowledge elements.

AUTOCART at process level model, portrayed in

Figure 7, provides an architectural view of the

anticipated processing for generation of Knowledge

Nodes, mainly through links – obtained from the

filtering process – and latent data semantics as

determined by specification and categorization of

the input data, from designated data streams.

Figure 8: iMap (intermediate state diagram), illustration

through Kohonen Nets.

22 23

32 31

t.c. 1

t.c. 2

t.c. 5

t.c. 3

t.c. 6

t.c. 4

Knowledge

nodes

Textual Data

Feeds

Textual

Content

Hyperlinks

Textual

Content

Annotation

Association

Links

Classification

Links

Relationship

Links

Data

Semantics

- What

- Who

- Where

- When

- Why

- How

Text Category

refine to

contains

Knowledg

e node

relate to

Knowledg

e node

assign to

ICEIS 2006 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

354

Figure 8 demonstrates the modelling of knowledge

nodes generation, after textual content (t.c. = 1…n)

have been categorized, following method given by

Kohonen Nets, for instance (Kohonen, 1990). The

categories of text produced by AUTOCART, in a

way illustrated in Figure 8, forms the core of the

knowledge nodes, accompanied by reference

information such as extracted documented

experience within the organisation, related

communities of practice and referenced expertise.

This enhanced structure serves our purpose, which

is not only the administration of electronically

available information, but also a viable

representation of the intellectual environment

aiming to make information actionable and relevant

within contexts of expertise coverage. Put simply,

we aim to combine all valuable reference

information in a framework to which everyone can

relate to, effectively leveraging the organisational

intellectual assets. These knowledge nodes shall be

of little value unless presented in an illustrative

form. Therefore, it was chosen to generate

cartograms to reflect knowledge instances

comprising such nodes. Our approach is to be

heavily based on the concept of self-organising

maps (SOM). Predefined text categories, either

domain or project or user specific, play the role of

input vectors while knowledge nodes correspond to

neurons. The main concept behind this analogy is to

place the winner topologically in the text categories

space, according to its relevance for containment of

the surrounding text categories. Figure 8 is

representative of the intermediate step of this

approach.

5 CONCLUSIONS AND

OUTLOOK

It is believed that automated organizational

cartography and knowledge modeling with a

computationally assisted model, inline with

considerations for its evolution studies, particularly

focusing on utilizing inputs and outputs of the

processes in strategic decision-making. Would

inevitably lead to a creation of an environment by

which organizational intelligence and innovation

spirals. Technology is symbiotic with what it is

conducive of, how are such data feeds provoked to

process content, would permeate for its utilization.

Consequently inferencing based on what the

semantics of knowledge withhold.

REFERENCES

Boisot, M. (1987) Information and organizations: The

Manager as Anthropologist, Fontana/Collins, London.

Booch, G. Rumbaugh, J. Jacobson, I. (1999) “The Unified

Modelling Language User Guide”, Boston: Addison

Wesley.

Hedlund, G. (1994) “A model of knowledge management

and the N-form corporation”, Strategic Management

Journal, 15, 73-90

Kohonen, T. "The Self-Organizing Map" Proceedings of

the IEEE, Volume 78, Number 9. September 1990. pp.

1464-1480.

McLoughlin, H. Thorpe, R. (1993) “Action learning – a

paradigm in emergence: the problems facing a

challenge to traditional management education and

development”, British Journal of Management.

Nonaka, I. Takeuchi, K. (1995) The Knowledge Creating

Company: How Japanese Companies create the

Dynamics of Innovation, Oxford University Press,

Oxford.

Nonaka, I. Umemoto, K. Senoo, D. (1996) “From

information processing to knowledge creation: a

paradigm shift in business management”, Technology

in Society, 18(2) 203-18.

LOCATING KNOWLEDGE THROUGH AUTOMATED ORGANIZATIONAL CARTOGRAPHY [AUTOCART]

355