Abstract
We present a formalism and a method that allow to learn structured representations in a noisy knowledge base. The formalism fills the gap between Artificial Intelligence and Data Analysis research's domains. It describes a kind of structured modal object called “hoard”, that can express various semantics as probability, possibility and belief. The method aims at growing and refining a knowledge base, composed of hoards, by the incremental use of a data base, composed of hoards subparts. Two levels of knowledge are involved: structured hoards represent classes, and non structured individuals represent data. The main goals of the method are to match quickly both levels by automatic rules generation, and to acquire new knowledge in the structured level, by using the flat data base. Some applications are, for example, situation understanding, image analysis, adaptation to temporal variable process, negotiation.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call