Computer-based interaction analysis (IA) is an automatic process that aims at understanding a computer-mediated activity. In a CSCL system, computer-based IA can provide information directly to learners for self-assessment and regulation and to tutors for coaching support. This article proposes a customizable computer-based IA approach for a generic synchronous CSCL system, that is, a system that can be customized for different learning tasks and different ways of performing these tasks. In a generic system, IA mechanisms must also be generic. In the proposed system, called Omega+, a specific submodel specifies the properties and rules that customize the IA process for the learning situation defined by three other submodels (process, interaction, and artifact submodels) that all parameterize the generic kernel. The feasibility of such a generic model-based IA approach is assessed, together with the efficiency of some of its underlying mechanisms and its global acceptance by users. (Contains 3 tables and 4 figures.)