Intelligent system and solutions for knowledge management in virtual research teams--ontology and expertise map.

Draghici, Anca ; Molcho, Gila ; Draghici, George 等

1. INTRODUCTION

Misunderstandings between distributed team members and faulty translations of software applications contribute to the rising costs of interoperability in virtual, distributed organizations. Indeed, the growing implementation of distributed software agents necessitates developing and adopting a shared terminology and syntax for efficient and effective interoperability. Ontology offers a solution for solving the interoperability problems brought about by semantic obstacles, that is, obstacles related to definitions of business and scientific terms and software classes. Ontology is taxonomy of concepts and their definitions supported by a logical theory. It is often captured in the form of a semantic network--a graph whose nodes are concepts or individual objects and whose arcs represent relationships or associations among the concepts (Huhns & Singh, 1997). Ontologies may differ not only in their content but also in their structure and implementation. Various methodologies exist to guide the theoretical approach taken, and numerous ontology-building tools are available. The problem is that these procedures have not coalesced into popular development styles or protocols, and the tools have not yet matured as in other software practices. However, ontology is typically built in more or less the following manner (Denny, 2002): acquire domain knowledge; organize the ontology; check the work; commit to the ontology.

Based on these considerations and perspectives, the present paper will outline the detailed approach used to build the ontology and complementary knowledge map for a particular virtual organization, the Virtual Research Laboratory for a Knowledge Community in Production (VRL-KCiP), a Network of Excellence (NoE) established in the context of the 6th Framework Programme (www.vrl-kcip.org).

2. ONTOLOGY IN THE VRL-KCIP NoE

The central aim of the VRL-KCiP NoE is to create synergy by integrating the research expertise and capabilities of the different member teams to support product life cycle engineering in the modern manufacturing environment. Hence, knowledge sharing and collaborative research constitute the core competency and potential for the network's success and the essence of its existence. Among the central activities required to fulfill the VRL-KCiP vision are: (a) building an ontology with the purpose of generating a common reference language among the member teams that can overcome differences in culture, location, language, and fields of expertise; (b) implementing a central knowledge management system (KMS) that will allow expertise identification and knowledge-sharing capabilities; (c) implementing IT-enabled one-to-one or many-tomany communications capabilities to complement the face-to-face meetings of the distributed network. Because of the potential major impact of ontology development on the success of the network, significant effort was invested in completing this task efficiently and effectively.

Ontology-building focuses on what the ontology is required for (Gruber, 1993). The VRL-KCiP ontology was developed to enable knowledge sharing and reuse. Initially, the ontology had two objectives: (1) to ensure a common understanding of specific terms describing members' fields of expertise and research relevant to state-of-the-art life cycle engineering; (2) to provide the structure of the VRL-KCiP knowledge map. The goal of the knowledge map was to enable explicit charting of member expertise to clearly define and locate experts within the network and to develop a concise core competency depiction (Molcho, 2008).

The need for the VRL-KCiP ontology was magnified by the nature of the network--a virtual multilingual, multidisciplinary, multicultural dispersed research team, researching the state-ofthe-art in the vast field of life cycle engineering that, contrary to most virtual enterprises, did not evolve gradually from a central core but rather emerged as a fait accompli.

During the process of developing the ontology it rapidly became evident that in addition to the points outlined above, the ontology would provide the structured context required to cultivate a high quality knowledge base for capturing, accessing, archiving, and validating knowledge objects in the VRL-KCiP knowledge management system (KMS). The following discussion focuses on the main stages in achieving the above goals.

2.1 Goal and methodology definition

Ontology definition is an art. Therefore, compromises had to be made. As a result, although the goals of building the ontology were clearly defined in the initial O stages of the network of excellence (NoE), opinions differed regarding which methods should be used to best achieve the goals, and many concerns were raised: (1) ontology construction is not yet well understood; (2) the size and complexity of the research domain is large; therefore, care must be taken to clearly define the scope; (3) there is no single correct methodology for ontology building.

In July 2004, a special knowledge management working group met in Troyes, France to establish vertical integration among tasks related to knowledge and to begin creating the network's ontology. A bottom-up approach was adopted, based on input from network members. An initial brainstorming session was held to establish the basic structure and instances of the VRL ontology.

2.2 Developing the ontology

A top-down approach was applied to define the ontology structure and determine its initial levels. At the highest level member expertise were divided into: (a) life cycle-related knowledge and (b) product-specific knowledge. At the next level, the life cycle-related knowledge was further detailed to specific life cycle stages (Design, Manufacture, Service, and End-of-Life (EOL)). These life cycle stages were then divided into substages. Next, emphasis was placed on collecting (1) approaches, (2) methods, and (3) tools. A bottom-up approach was then applied to explicate further levels of detail and gather instances and documents for each type of expertise. The ontology--both structure and content--was then further developed through iterative steps of collecting, analyzing, brainstorming, revising, and redistributing for further feedback. This process continued until a relatively stable ontology structure was formulated (Van Heijst, et. al., 1997).

2.3 Collecting expertise profile

Once the ontological structure was more or less defined and stabilized, the form was again distributed to all VRL-KCiP team members. To date, 250 responses have been received and entered into a collective knowledge base. Many new instances have been added to the basic structure, as members sought to define their personal expertise. The first stage of expertise collecting did not incorporate differential rating of personal expertise (Molcho, 2008).

2.4 Creating the VRL-KCiP knowledge maps

In accordance with the ontology structure and the responses from network members, all feedback was entered into a common database. This database made possible to map the expertise of the individual network members as well as to combine the input from individuals from each lab with the fields of expertise available in each lab. Four knowledge maps have been built: (a) individual expertise range, (b) individual expertise, (c) lab expertise strengths, (d) lab expertise. They were built by assigning the value 1 to all expertise fields relevant to each network member (Molcho, 2008).

2.5 Completing ontology instance profiles

To give the VRL ontology added value as well as added dimension and depth, a profile structure was defined for each instance. The profile for each field of expertise aids members in understanding the context of each instance. It also provides a basic introduction to the topic. Work is currently underway to complete and validate a profile for all instances in the ontology. Over 150 profiles have already been incorporated in the VRL-KCiP KMS. Each profile provides a definition of the instance, a short description, and good references to further information. It also provides a link to a more detailed description that outlines strengths, weaknesses, complementary tools, applications, and so forth.

2.6 Ongoing ontology updating

Since ultimately there is no correct ontological structure (each proposition has its benefits and drawbacks) and since a platform must be in place to initialize joint ventures and research, we have refrained from major changes in the structure. Nevertheless, the ontology continues to evolve for a number of reasons. The first reason for ontology evolution is determined by the upper part of the tree (product life cycle). The expertise maps indicate a lack of balance between the level of detail of the design phase, which is the most explicit, and the service and EOL phases, which lack detail. This imbalance appears to mirror the fact that the strength of the VRL-KCiP lies in the design phase (design approaches, methods, and tools), whereas the network lacks expertise in the service and EOL phases so that the structure is sparsely populated in these areas. More effort must be invested in further detailing the service and EOL product life cycle stages. One of the ways to further detail these areas is by applying bottom-up methodologies more common in ontology development (topic mapping or text mining methodologies). This work, currently underway, involves mining member CVs and current areas of research being collected on the central KMS (Draghici & Draghici, 2005). The second reason for ontology evolution is determined by the lower part of the tree (products section); the products section of the tree will be built from the bottom up, and branches are likely to be added as new members with expertise in specific product types join the network. New instances are added to the ontology as new members join the network, new fields of research evolve, and research projects begin. Hence, the bottom-up process of expanding the tree to include new fields of research relevant to the network and new tools or methodologies developed within the labs will continue. The ontological structure will expand further, both in depth (further detailing of existing branches) and in breadth (by introduction of additional fields of expertise).

2.7 Expertise level differentiation

On the first competence profile, some members filled in only those instances in which they were very highly knowledgeable, whereas others filled in all the instances in which they had basic knowledge. This discrepancy, apparently due to participant personality differences, resulted in an unbalanced picture for understanding lab capabilities. Hence, a differential rating was required. For each direct instance (leaf in the tree structure), the user can indicate his/her appropriate level of expertise (familiar, novice user, experienced user or teacher, and innovator or developer). Based on these results, the process of expertise differentiation will continue lab by lab until the level of expertise is mapped as well.

3. CONCLUSIONS

After almost two years of development, the ontology currently fulfills four major purposes in the network: (1) as a reference for a common understanding of terms in the fields of research relevant to the VRL-KCiP; (2) for collaboration definition and initiation; (3) as one of the indexes for the dualindex KMS; and (4) as the coordinates of the VRL-KCiP knowledge map describing the current expertise of each member in the network, thus representing its intellectual capability and core competence. Although the VRL-KCiP KMS is the core of the EMIRAcle association development (it will be continue up-date and enriche) as a fesable partner for industrial solutions in research and innovation (www.vrl-kcip.com).

4. REFERENCES

Denny, M. (2002). Ontology building: A survey of editing tools. Retrieved November 2, 2007, Available from: http:// www.xml.com/pub/a/2002/11/06/ontologies.html, Accessed: 2008-05-13.

Draghici, A., & Draghici, G. (2005, November). Building a knowledge share culture in VRL-KCiP NoE. Presentation at the VRL-KCiP Knowledge Management Workshop, Nantes, France.

Gruber, T.R. (1993). A translation approach to portable ontologies. Knowledge Acquisition, 5(2), 199-220.

Huhns, M.N., & Singh, M.P. (1997). Ontologies for agents. IEEE-Internet Computing, 1(6), 81-83.

Molcho, G. (2008). Formulating an expertise map in the VRLKCiP. In Bernard, A., Tichkiewitch, S. (Eds.), Methods and Tools for Effective Knowledge Life-Cycle-Management, Springer, pp. 169-184.

Van Heijst, G., Schreiber, A., & Wielinga B. (1997). Using explicit ontologies in KBS development. International Journal of Human Computer Studies, 46, 22-35.