Usages Analysis in Instant Interpersonal Communications over IP

This paper investigates the contribution of data mining techniques in order to optimize IP based real time services such as instant messaging or telephony over IP. The main difficulty is here to handle temporal constraints and to interpret the corresponding raw data into users´ behaviour high level knowledge. As an example, we focus on IM and try to find a light mean to detect conversations consistent from the human point of view (i.e sum of several basic exchanges)