Scientific and technical knowledge capitalisation at IRSN
Helmut PITSCH K. Gérald FAVRE Sylviane BONNEFOUS
IRSN, Direction scientifique FAVRE Consulting SARL IRSN, Direction scientifique
BP N°17 1, Allée de Coubron BP N°17
92262 Fontenay aux Roses, France 93390 Clichy sous Bois, France 92262 Fontenay aux Roses, France
+ 33 1 58 35 91 45 + 33 6 13 08 07 27 + 33 1 58 35 76 62
helmut.pitsch@irsn.fr gerald.favre@favre-consulting.net sylviane.bonnefous@irsn.fr
choosing the best editor-integrator team. Intense testing, precise
1. INTRODUCTION specifications and a rather extended pilot-phase made it possible
The poster points out the most relevant aspects of the CCST to open a system without any but cosmetic bugs.
project, which led to recently open the Scientific and technical
knowledge capitalization system at IRSN (Institut de
ion
radioprotection et sûreté nucléaire): a federation, by a full-text Cr fus
aw Dif
lin
research and analysis engine, of several databases where the g
Institute’s scientific documents are stored. Emphasis is put on Knowledge transfer
Brevets Veille stratégique Transfert du
Intelligence
savoir AT
IO
N
Complementary
Knowledge
Cartographie
mapping
du savoir IS
original need, modelling of knowledge management, decision- Communauté de pratique
Community practiceGPEC EC
I AL
sub-systems
SP
making process, project method and final result, including first Knowledge
Gestion du
and
savoir
know-how
et du savoir
management
-faire
Accumulation
comments on acceptance by end-users. N
Valorisation
Exploration
Text mining de texte T IO Central
Identification U RA
Identification of
desexperts
experts de concepts
of concepts CT pivot
RU
ST
2. FROM THE NEED TO THE TOOL Capitalisation
Knowledgedescapitalisation
connaissances
The institutional context includes the specific role of IRSN as the Archives
Courrier
Mail Archives S
French technical support organization in the nuclear field, its UR
CE
Decentralised
Documents Ouvrages
Books
SSO
main characteristics and organization, and the consequent need to RE management
Electronic documents et
management
reorganise internal and external knowledge fluxes on a share-of- Gestion de ’linformation des documents
information basis in order to improve accessibility and exchange.
ing
ed
Fe Concept : FAVRE Consulting
An initial diagnosis phase was dedicated to identify existing
habits and tools, needs for new methods as well as cultural and
technical constraints. The results of the inquiry were used to build
up a model for amplified knowledge capital accumulation (Figure Figure 1: Model for amplified knowledge capital accumulation.
1) and propose a knowledge management strategy to improve the An advanced syntactic research engine (AMI Enterprise
daily used tools and create new habits: centralized data Intelligence) from AMI Software [1] forms the heart of the system
management was rejected because of the gigantic dimension of and enables any habilitated employee (about 2200) to explore ten
the required system, the limits of the maintenance capability, as databases where the different specialized departments store their
well as the historically anchored culture of developing specific papers, posters, reports, technical notes, presentations etc…;
tools adapted to local needs. So it was decided to develop transverse databases (quality reference documents, archives…) are
logically circumscribed and technically compatible sub-systems, also included. All interfaces development and integration work
adapted to each function (intelligence, business management, was carried out by Bull SAS [2]. As every database has its own
storage…) and up-to-date the existing tools that needed to, with technology and functional management, every interface requires a
the aim to put them progressively together into a coherent system. specific development in order to properly transcribe the metadata
Centralized tools should be dedicated essentially to distribute and in a homogenized format and especially to respect all specific
analyse structured information. restrictions in the access to information imposed by the different
The knowledge capitalisation sub-system is dedicated to explore divisions of the Institute.
and analyse the internal scientific and technical production of the The system also allows statistical analysis and graphic
Institute: a research engine was connected to existing databases, representation of results by date, origin, type of document…, a
the management of which remains in the operational departments. useful tool for managers as well as for researchers. A special
All along the project, priority was given to functional aspects, modulus is dedicated to concept extraction and knowledge
end-users were associated to every reflexion and test phases, an mapping, in order to detect experts or evolutions in research
external field expert was present during the whole process and domains for example: automatic cluster characterization allows
technical developments were sub-contracted. As a basis for the differentiating strong and weak signals in a large corpus of results.
call for tender, a state-of-the-art study was carried out in order to
develop an internal knowledge on automatic language analysis Regarding main characteristics of the system, priority is given to
and evaluate the performance of different engines. A specific form precise information recovery (relevancy is the first criterion),
of call of tender which includes a competitive dialogue allowed extension of requests with a specialised thesaurus [3], reliable
statistical analysis and mapping of results. It gives access to about
180 000 indexes and 20 000 documents. These figures will After three months of using, it is impossible to conclude on
duplicate during the current year after connecting three new bases. acceptance, effective users are not yet representative of the whole
As previously mentioned, particular attention is paid to security of Institute, but comments are positive and critical remarks are
access to information. User recognition is daily refreshed by oriented towards improvement suggestions. Reserves towards
synchronisation with the Institute’s directory and confidentiality sharing information are progressively diminishing and the
rules to access information are retaken specifically from each structuring function of the system may induce knowledge
database. capitalisation in different ways. Future developments will include
connexion with the Institute’s intelligence system used to retrieve
Basic functions are quite intuitive and easily dominated through
and analyse scientific information from external databases.
self-learning; about 1% of end-users received formal teaching,
they also participated to the testing process and are in charge of
helping colleagues in mastering the more specialised aspects. 3. REFERENCES
Much lighter presentations are periodically made at the different [1] http://www.amisw.com/fr/produits/ami-entreprise-
locations of the Institute to direct the users and induce them to intelligence/fonctionnalites.htm
take advantage of the most powerful possibilities of the system [2] http://www.bull.com/fr/services/integration.php
with the aid of a specifically designed User’s guide. Specific
questions are answered by the administrator through a dedicated [3] IAEA / DNE / INIS & NKM Section. April 2007.
E-mail box. ETDE/INIS Joint Reference Series No. 1 (Rev. 2).