Intelligent Environments are expected to act proactively, anticipating the user's needs and preferences. To do that, the environment must somehow obtain knowledge of those need and preferences, but unlike current computing systems, in Intelligent Environments the user ideally should be released from the burden of providing information or programming any device as much as possible. Therefore, automated learning of a user's most common behaviors becomes an important step towards allowing an\udenvironment to provide highly personalized services.\udIn this paper we present a system that takes information collected by sensors as a starting point, and then discovers frequent relationships between actions carried out\udby the user. The algorithm developed to discover such patterns is supported by a language to represent those patterns and a system of interaction which provides the\uduser the option to fine tune their preferences in a natural way, just by speaking to the system.
展开▼