The proliferation of data produced by sensors and IoT technologies drives a new level of user data exploitation. Analyzing those data enables an intelligent understanding of user behavior and context to predict user intention correctly and provide relevant service recommendations. Many studies have exploited data from different sources for process mining purposes, but a few covered intention mining and its relation to context data. In this paper, we propose an architecture that uses information from data emissions to extract the user’s situation, recognize his intention and context and then propose relevant services to the user according to his intention. In addition, our approach includes lightweight ontology to script the semantic annotations of the data stream.