Researcher's speakingCapitalizing on company knowledge, either explicit or tacit, poses human, organizational, economic and technical problems and requires a multidisciplinary approach. An ideal corporate memory must supply the right person with the right knowledge at the right time to make the right decision. Several INRIA teams are taking up this challenge by working on three types of memories, technical memory (Acacia and Sherpa), profession memory (Acacia) and project memory (Acacia).
 |
| Automobile project memory |
The life cycle of corporate memory can be broken down into several phases: needs detection, building, distribution, utilization, evaluation and evolution of the memory. Acacia Project proposes a methodological guide to analyze the whole cycle. The Eiffel team at INRIA Rocquencourt has defined cognitive models of a designer in a multi-professions industrial situation, for the detection of needs. Building up a memory often requires several sources of expertise. Acacia thus proposed models to manage the diverse points of view of several experts and a model for concurrent design tasks. Sherpa Project proposed a protocol for collaborative building of consensual artifacts (CO4). Memory is materialized by documents and knowledge bases in Acacia and Sherpa. The interest of representing knowledge in an explicit and formal fashion is to develop such knowledge formally and be in a position to apply techniques related to the formal structure. In effect, a formal language makes various verifications and operations possible. Since not everything can be expressed in a formal way, it is necessary to link this knowledge with such informal sources as texts, bibliographies, lexicons and images. Depending on the context, a company memory user may be interested in the informal sources (texts) or the formal sources (knowledge bases). Acacia and Sherpa have both adopted this approach, but used different knowledge representation formalisms: CommonKADS and conceptual graphs for Acacia and objects for Sherpa.
Memory is then spread by a knowledge server through the company's Intranet or the Web. Acacia developed WebCOKACE, a knowledge server capable of distributing CommonKADS expertise models on the Web. Sherpa developed Troeps, a knowledge server distributing object knowledge bases. Information search in the memory can be guided by the knowledge models. Acacia is thus studying information search in XML documents annotated by RDF (Resource Description Framework) knowledge models or by conceptual graphs. These methods and software (Troeps and WebCOKACE) gave rise to various industrial applications in accident analysis with Inrets, aeronautics with Aérospatiale and Dassault-Aviation, car manufacturing with Renault and Peugeot and in microelectronics with ST Microelectronics and Dassault Systèmes.
|
| Cokace: a CommonKADS expertise model Web server |
