Indexing Corporate Memories through Ontologies
Djida Bahloul, Youssef Amghar, Pierre Maret
INSA de Lyon – LISI, 7 avenue Jean Capelle.
69621, Villeurbanne – France.
dbahloul@lisi.insa-lyon.fr, {amghar,maret}@insa-lyon.fr
Abstract ontological relationships. This is followed in section 2
In the context of Knowledge Management, we carry out by a description of our model called S3 allowing both
a Corporate Memories (CM) project for the Company representation and indexing of knowledge. In the
CIRTIL1. Our purpose is to focus on the modelling of section 3, we apply this model to build an ontology and
the application domain. It is built as a domain ontology to index the CM of a company.
with a structure supporting a semantic model based on
ontological relationships. In this paper we, present our 2. RELATED WORKS AND POSITIONING
S3 model which permits to model knowledge and to Ontologies can constitute a component of a CM: they
index documents. We also show how semantic indexing can be explored by the end-user to discover the
of technical documents can be improved by mean of the organization processes and business objects of the
domain ontology. We show finally the interest of our enterprise (e.g. enterprise ontology), or to study
model with the implementation of a prototype. methodically a specific technical domain (e.g. domain
ontology), etc. Ontologies are then used as a coherent
Keywords support to describe and to share knowledge. “Ontologies
Semantic indexing, ontology, corporate memory, constitute the glue that binds knowledge subprocesses
knowledge modelling. together. Ontologies open the way to move from a
document-oriented view of Knowledge Management to a
1. INTRODUCTION content-oriented view, where knowledge items are
Corporate Memories (CM) require abilities to manage interlinked, combined, and used.» [Staab and al., 2001].
disparate information and heterogeneous sources in In fact, ontologies are used more and more in KBS
order to make knowledge accessible to the adequate (Knowledge Management System) development. For
users of the enterprise. CM must also consider the example, projects such as SHOE [Heflin and Hendler,
integration and the storage of knowledge contained into 2000] and Ontobroker [Benjamins and Fensel, 1998]
electronic documents or contained into knowledge use ontologies to improve the searching abilities on the
bases. According to us, to permit the success of using World Wide Web. Both systems provide logical
and maintaining the CM, it is important to consider reasoning based on ontological definitions. In [Gandon,
employees as the hard core of the system. Indeed, 2001]; [Gandon and al, 2002], CoMMA project offers a
employees can participate to capture knowledge and to solution to implement a CM based on ontologies and
structure it into the CM. The objective of our research agent technology. It promotes a wide vision of the
works aims to offer to users a methodological assistance document retrieval issue that could be applied to several
and tools enabling the knowledge management. To cases. The memory is composed of heterogeneous
achieve this objective, our approach takes advantage of evolving documents, structured using semantic
both: the contribution of ontologies as proposed by the annotations expressed with concepts and relationships
Artificial Intelligence (AI) community and the provided by a shared ontology. In others approaches,
documentary indexing such as defined in the domain of ontologies are exploited to organize the knowledge and
the library science. The role of ontology is the to support computational design. For instance, the
representation and the modelling of knowledge. “An approach for ontology-based knowledge management
ontology is an explicit specification of a [Staab and al., 2001] includes a tool suite and
conceptualisation” [Gruber, 1993]. The role of methodology for developing ontology-based Knowledge
documentary indexing is the matching between the Management systems. Among these tools,
represented knowledge and users’ queries. In this paper, OntoAnnotate tool allows users to create objects and
we propose an indexing model which is more efficient describe them with their attributes and relationships.
than a simple taxonomy of concepts. Therefore, we This outline of the state of art shows that modelling
build a domain ontology which has a significant based-ontology is interesting in the frame of CM.
capacity of expression thanks to the possibility of Nevertheless, the difference between our approach and
introducing semantic links, structural links and related works consists in using ontologies to carry out
subsumption links between concepts. We show in this the documentary indexing of formalized knowledge. We
paper how to facilitate the indexing of technical understand by documentary indexing, "the operation
documents through the ontology. We outline our which consists to describe and to characterize contents
approach for the construction of an ontology based on of documents by using representative concepts." In our
approach, this operation is carried out by the ontology
1
(instead of a documentary language). We argue that
CIRTIL : Centre d’Informatique Régional du Traitement de
l’Information Lyonnais which supports this work.
ontologies can guarantee a sufficient indexing because,
they introduce a host of structural and conceptual
relationships including super class/subclass/instance This complexity requires an indexing model that is most
relationships, property values time relationships, and possible representative of the domain.
others depending on the used language representation
[Saadani and Bertrand-Gastaldy, 2000]. In addition, an 3.2 An Indexing Model on Ontological Relationships:
ontology can also infer that one concept is a special case The S3 model
of another because the logical definition of each concept To elaborate the S3 model, we first built an ontology
can be compared. If the concept C2 satisfies the called OntoCIRTIL that intends to model technical
requirements being a specialization of concept C1, then incidents, employees who treat these incidents, products
C2 can automatically be classified below C1. We concerned by these incidents and the entities required
consider thus, the ontology adequate to allow indexing for the characterisation of these incidents, products and
knowledge. employees. The particularity of OntoCIRTIL is to model
and to represent knowledge according to various facets
3. TOWARDS AN INDEXING APPROACH and viewpoints by taking advantage of conceptual
BASED ON ONTOLOGY relationships. The S3 model proposed in this paper is
based on three views giving three spaces for organizing
3.1 Complexity of the domain of knowledge
semantic of knowledge fragments represented by
Our approach is proposed in the frame of the design of a
concept of domain ontology. These three spaces are
CM for the company CIRTIL. This company aims to
described below:
save and capitalize its knowledge and its know-how
1. Semantic space: This space gives a view allowing
concerning the production activities related to technical
discovering concepts through semantic links. This space
incidents on database servers and applications. We aim
is domain dependent because relationships used here are
at providing users with support tools for indexing
defined and interpreted by users. Our environment
knowledge especially contained into technical
offers primitives to define such relationships such as:
documents. Knowledge of this company has many
create, delete, and rename …a semantic link. The
senses according to the context (administrator of
graphs path algorithms make possible to navigate in this
database vs. administrator of administrative
space in order to discover concepts: to displace from a
department). As well, synonymous terms are used in
concept to another according to the semantic link
different situations (application vs. software), and words
defined by the expert. This space embeds semantic
have a large variety of different links and
links.
interrelationships (report is a kind of documents and
2. Structural space: This space makes proposes to
chapter is a part of a document). This requires a model
structure knowledge considering the structural
based on a formal domain ontology which considers all
dimension of the knowledge fragments. These
those characteristics. The interest of such model relies
fragments are represented by concepts related according
in the role of these relationships in the semantic
to aggregation links (derived from the structuring
representation. We studied some ontological
mechanism of the Object-Oriented Approach). Besides
relationships composing ontologies and we concluded
the graphs path algorithms, it is interesting to define a
that these relationships define and enrich the semantic
zooming function. Its goal is to discover various levels
between the concepts. Indeed, the type of relationship
of the selected concepts. This zooming function allows
can change the semantic as showed in the following
exploring deeply the various parts composed one
example. Thus, two concepts C1 and C2 linked by two
concept. This space allows representing structural links.
distinct relationships R1and R2, give two different
3. Subsumption space: This space makes it possible to
semantic:
organize the definitions of concepts by using
R1(C1,C2) gives S1
subsumption links. This space increases semantic space
R2(C1,C2) gives S2
by using a new dimension which allows improving
Let us consider two concepts “Incident” and
concepts in term of related definitions. The graphs path
“Employee” and two relationships “reported-by” and
algorithms will make it possible to find a concept about
“treated-by” (Figure 1). Semantic can be variable
a large definition. In fact, this space allows organization
according to the relationship between each of these two
of knowledge in manner to permit employees to retrieve
concepts. The meaning of the first representation is
concepts (and related documents) from definition,
different from that of the second representation. Thus,
which concerns their ancestors in the subsumption
Reported-by (Incident, Employee) means that a given
graphs. This space includes subsumption links.
incident has been notified by an employee during his
task; and Treated_by (Incident, Employee) meaning that 3.3 Description of link types of S3
an employee has resolved the incident. In our model each concept is linked with one (or
several) other (s) concepts by at least a type of links:
Incident Reported-by Employee Semantic link, Structual link, Subsumption link. Each
type of relationships interprets a particular semantic
Incident
between two concepts. To model these relationships,
Treated-par Employee
our approach is based on manual linguistic analysis. We
applied a same method adopted to define the concepts,
to choose the representative links, i.e. we proposed
Figure 1 - Relationships for Semantic several lists, which were modified then validated by
employees according to the ontological commitment.
1. Semantic links: The main characteristic of this type control semantic relationships between concepts in
of relationship is such as the nominations of links order to support clarification reasoning. Inference link
related to the usual language (natural language) of the functionality is supported by inference engine, which
community for which ontology is available. Note that, uses concepts of ontology to derive factual knowledge
representation of semantic links, it means the that is only provided implicitly. The inference engine is
nomination of relations, is not standardized. It varies used to derive information that is implicitly present in
according to each designer. It can be represented in: resources of knowledge without requiring that all
verbs or prepositions [Sherratt and Schlabatt, 1990]; information is complete materialized by relationships.
verbs or nouns [Heeren and Collis, 1993]; verbs or Unfortunately, the inference link which is very
logical connectors [Malone and Dekkers, 1984]. This important in the context of information retrieval, is not
facilitates their use in especially for retrieving present in the thesaurus.
knowledge in the CM because, these relationships are
considered as keywords. Semantic relationships express 4. EXPERIMENTATION
clearly "evident" and no ambiguous knowledge. This part focuses on the Knowledge Base and indexing
Examples of semantic links used in our ontology are: tool “KnowIndex”. We explain how we partially
treated-by, written, organized, causes, necessitates, implement structural space of the model S3; then we
implicates. Note that those relationships allow the bi- present the mechanism of indexing process.
directional and opposite (inverse) semantic expression.
2. Aggregation links: Structural links constitute a kind 4.1 The Knowledge Base
of relationships, which expresses a strong property Knowledge about technical dysfunction (failure server,
between the whole and the parts, as well as unavailable application, defective networks, etc.) and
subordination between the existence of the parts and the documents of resolution of these incidents represent a
whole. The parts associated to a non fixed multiplicity category of crucial knowledge for the company.
can be created after the composite itself, but once Employees and decision-makers need to capitalize this
created, they "lives" and "dies" with him (i.e. they share knowledge, within sight of access and re-use this
its duration life). Such parts can also be explicitly knowledge in order to improve quality of services for
withdrawn before the death of the composite. The customers. In order to contribute to build an
composition can be recursive. This type of relationship environment for CM we developed a tool for hitches
represents links of dependence between concepts. The management called MaTIP for "Management of the
choice of the nomination of the links ("Part-of", Technical Incident Project". The main objective is to
"component-of") depends on the meaning that we wish capitalize data, information and knowledge allowing, as
to represent. For example, to express that the well as, the identification, the management and the
Functional incident and the Technical incident belong anticipation of dysfunctions and technical anomalies
to the Incident Family, the link component-of is better that occur during the exploitation time of the software
adapted than part-of. applications by customers. This Knowledge Base
3. Subsumption links: supports the ontology OntoCIRTIL (concepts and
In the simplest case, ontology describes a hierarchy of relationships identified during the conceptualization
concepts related by subsumption relationships. This process). Validated concepts and relationships are
relationship is largely used and it is considered as the modelled by the class UML diagram. In class diagram
foundation of ontology, because all ontologies are of UML, the concepts are represented by class and
presented in taxonomies [Guarino, 1997]. In such relationships by arcs. The model formalized by UML is
taxonomy more general concepts subsume more specific translated during this stage of the development of the
concepts. This allows information about concepts to be Ontology into RDF. Compared to existing knowledge
associated with their most general concept, and it allows representation standards, e.g., KIF [Genesereth and
information to filter down to more specific concepts in Fikes, 1992], KRSL [Allen and Lehrer, 1992], or
the taxonomy via inheritance. To name the relationship, DAML+OIL [Broekstra and al., 2001], RDF can be
we can use the "is-a" link or "kind-of" according to the considered as a not very expressive formalism.
adequate semantic of the context. Formally, an “is-a” Nevertheless, it offers an adequate expressiveness for
link is an implicit subset/superset relationship between our application needs. RDF uses a simple data model
the two concepts. A simple example of this relationship expressed in XML syntax to represent properties of
is the Incident is-a Event. Note that, this relationship resources and their relationships. The main element of
characterizes the hierarchical relationships between the RDF is the notion of rdfsResource. This one is at the top
key words into thesaurus. of the class hierarchy and is subclass of itself. A
rdfsResource has a rdfstype relationship with a
In addition of these three types of conceptual rdfsClass, and a rdfslabel relationship with a
relationships, our indexing model proposes a rdfsLitteral and may also have some literals as
mechanism, which allows deducting new knowledge in comments. One of the resources in our model is
order to enrich a semantic. In some case, knowledge is "Relation". It is a super class for Structural link,
also defined with rules: a Chief Project Manager is a Semantic link and Subsumption link. Also, each type
person who manages project. These rules permit to of these links has a value which is its object. For
express implicit useful knowledge of information example, linked is the value of the resource "Relation".
retrieval. We write some constraints that permit to
Figure 2 presents an example of the representation of 3. Indexing and Classification: This step is based on the
the relationships and its properties. conceptual model. Indexing task is initiated by the user,
thanks to the function “Add” associated to each
concept. By clicking “Add” on one of top-level
concepts or the sub-concept, “KnowIndex” creates
concept with the indexed document, the system supports
an automatic mechanism which permits the linguistic
analysis.
3.3 Knowledge Retrieval
To show the double role of the ontology OntoCIRTIL,
i.e. the restitution of knowledge and its indexing, we
propose a single user interface. Ontology can be seen as
a semantic interface for accessing CM. For that, the
ontology appears on the left of the user interface in a
Figure 2: Example of a representation using RDF tree view representation. To simplify the indexing task
to the end-users, we associate to each concept two
3.2 Indexing tool operations “Add” and “Show” (see Figure 3). These
We designed a prototype based on Java and XML operations facilitate the correspondence between a
technologies to implement the indexing engine document (and part of document) and the appropriate
“KnowIndex” that supports the semantic S3 model. One concepts. As previously mentioned, thanks to the “Add”
of the objectives of our contribution is to integrate the operation, knowledge can be captured and classified
indexing operation into the daily activities of the into the CM. The “Show” operation is used to search
employees’ company. In fact, “KnowIndex” is an for knowledge. Selected text and their links address are
application that employees can use easily to index or to classified and stored into the knowledge base. The
retrieve formalized knowledge. The indexing technique choice of concepts is manual because it is achieved
can be achieved through the ontology considered as an directly by the end-user. This user, can be able to
indexing resource. The indexing mechanism is divided determine the representatives concepts while consulting
into three steps; we address each of them in the a document. Furthermore, the “KnowIndex” module
following paragraphs. allows the linguistic treatment, makes correspondence
1. Selection of knowledge: The selection of knowledge between selected text and concepts, and classifies
is achieved in the usual environment of employees’ indexed documents and their concepts in the knowledge
work, for instance during the use of Microsoft Word base. Thus, the integration of contents or news
application (it could be Lotus Notes or any traditional documents into the Knowledge Base can be relevant,
employees’ office tool. This selection concerns as well because it is ensured by the users for themselves.
as, all content of a document or only a part of a
document (paragraph, sentence,…). Once the text is
selected, “KnowIndex” generates automatically the
selected knowledge into an XML file.
2. Selection of representative's concept: This step
consists in choosing relevant concepts in the ontology in
order to associate them to the selected knowledge.
Notice that the employee himself achieves this delicate
phase. To help employees, we reserved the left side of
the “KnowIndex” interface to display the ontology.
Thus, users can select either node or leaf by navigating
in the tree of the ontology. We conserve the right side of
the interface to the “Editor”. Its role is to visualize for
each concept all knowledge identifying these concepts,
particularly: the definition of concept, properties and
relationships. When the user displays a concept, the
concept definition, super-concepts, sub-concepts, Figure 3 - “KnowIndex” : the indexing operation
relationships and instances of the concepts are displayed We observe on the right side of this figure a technical document
in the frame on the right. These characteristics and in Word format. The user locates by traversing the document a
information about concepts and relationships guide the paragraph describing a given dysfunction. To record this
user to choose a concept compared with another fragment of Knowledge, the user selects in a traditional way by
using the mouse. Then, he browses the ontology in the
concept. application "KnowIndex " in order to find, the representative
concept. Once the relevant concept is added, the system
automatically connects the concept to the paragraph.
5. CONCLUSION
The advantage of the ontology according to knowledge [Genesereth and Fikes, 1992] M. R. Genesereth and R.
engineering researches is that the concept must be E. Fikes (eds.). 1992. Knowledge Interchange Format,
unambiguous and unique. In addition, it is necessary Version 3.0 Reference Manual. Computer Science
and important to take advantage of the richness of the Department, Stanford University, Technical report
ontological relationships. This what guided our Logic-92-1.
approach which leads to the development of a semantic
model. In this paper we have. We presented the S3 [Gruber and Poli, 1993] Nicola Guarino and Roberto
model and its components. We explained that the Poli, Toward Principles for the Design of Ontologies
ontological relationships allow a strong semantic. In this Used for Knowledge Sharing, In Formal Ontology in
context, we proposed three type links, which we Conceptual Analysis and Knowledge Representation,
analyzed and modeled. The experimentation applied to a Kluwer Academic Publishers, in press. Substantial
project of CM permits first, to expose real needs and revision of paper presented at the International
then to test and validate our approach. Thanks to the Workshop on Formal Ontology, March, 1993, Padova,
indexing tool "KnowIndex" dedicated to assist actors, Italy.
we developed a process which, allows:
- the selection of knowledge; [Guarino, 1997] Guarino, N. 1997b. Understanding,
- the selection of representatives' concepts; Building, and Using Ontologies: A Commentary to
- and the Indexing and Classification. "Using Explicit Ontologies in KBS Development", by
van Heijst, Schreiber, and Wielinga. International
The advantage of our contribution is to develop an Journal of Human and Computer Studies (in press)(46):
indexing model which exploits the ontological (in press).
relationships in order to allow a level of abstraction and
a semantic expressiveness. Nevertheless, the major [Heflin and Hendler, 2000] Heflin, J. and Hendler, J.
problem of building our model is the semantic Semantic interoperability on the web. In Extreme
formalisation of the cognitive knowledge. This is Markup Languages 2000.
accentuated by the fact that the realization of the model
is time-consuming. The application of the model to a [Heeren and Collis, 1993] Heeren, E. and Collis, B.
small corpus showed that the approach based-ontology Design considerations for telecommunications-
can proven time-consuming in particular when the supported cooperative learning environments : concept
ontology must be built. Also, only the structural space mapping as a telecooperation suppport tool. Journal of
was formalized and implemented. Thus, our proposal educational multimedia and hypermedia, 4(2), 1993,
will become more interesting while we formalize and 107-127.
implement the semantic space and the subsumption
space and obviously by integrating the inference [Malone and Dekkers, 1984] Malone, J. and Dekkers,
mechanism. We also plan -in the future works- to J. The concept map as an aid to instruction in science
evaluate the relevance of indexing process based on the and mathematics. School science and mathematics,
ontology compared with that based on the thesaurus. 84(3), 1984, 220-232.
[Saadani and Bertrand-Gastaldy, 2000] Saadani.L,
6. REFERENCES Bertrand-Gastaldy.S. Conceptual Maps and Thesauri: A
Comparison of Two Models of Representation from
[Broekstra and al., 2001] D. Broekstra and al. Different Disciplinary Traditions CAIS 2000. Canadian
Enabling Knowledge Representation on the Web by Association for Information Science Proceedings of the
extending the RDF Schema. 10th International WWW 28th Annual Conference.
Conference, Hong Kong 2001.
[Staab and al, 2001] Staab S., Schnurr H.-P., Studer R.,
[Gandon, 2001] Gandon. F. Engineering an Ontology Sure Y.(2001): Knowledge Processes and Ontologies.
for a Multi-Agents Corporate Memory System In IEEE Intelligent Systems. 16(1), Special Issue on
proceedings of ISMICK 2001 Eighth International Knowledge Management, January/February 2001.
Symposium on the Management of Industrial and
Corporate Knowledge, 22-24 October 2001, Université [Sherratt and Schlabatt, 1990] Sherratt, C. S. and
de Technologie de Compiègne, France, pp 209-228. Schlabach, M. L. The application of concept mapping in
reference and information services. RQ, 30, 1990, 60-
[Gandon and al, 2002] Gandon.F, Dieng-Kuntz.R, 69.
Corby.O, Giboin.A. Semantic Web and Multi-Agents
Approach to Corporate Memory Management. 17th
IFIP World Computer Congress IIP Track - Intelligent
Information Processing, Eds Musen M., Neumann B.,
Studer R., p. 103-115, August 25-30, 2002, Montréal,
Canada.