Learner assessment in learning games (LG) is an interesting research area for both academia and industry. The play traces resulting from the learner’s activity in LGs with large state spaces and a large amount of free interactions, are hard to analyze and to interpret by teachers. In this paper, we present a framework to assist the building of an expert’s solving process that is the base of the algorithm that analyzes player’s traces and generates pedagogical labels about the learner’s behavior.