This article presents the formal bases of our complex object oriented language, together with our distributed architecture, originally developed for a speech recognition application, generalized as a tool that can be used regardless its field of application to conceive evolutive multi-expert bases of complex knowledge.

The originality in this system lies in the definition of a complex object oriented language CODEX (Complex Object Dynamic EXpert language). We have defined a complex object to represent a complex knowledge, that can be dynamically modified, is composed of parts and that eventually has temporal characteristics. This representation grants standardization, so that within the same system declarative and procedural knowledge derived from different experts can be brought into collaboration. Our system takes into account temporal aspects of reasoning and of knowledge representation. Standardization of knowledge and distributed control forecast an extension to our knowledge base system of Distributed Artificial Intelligence.