Computer-Supported Collaborative Learning (CSCL) is one of the most significant achievements leading to improve teaching and learning using information technology, as it seeks to activate the student-centered learning, in which the student is the main focus of the learning process. Chat is considered one of the most important CSCL tools which are used in knowledge transfer and information exchange. In fact, chat is an ideal tool that aims to realize the collaborative principle, which allows individuals to express their ideas and opinions through educational dialogues. We propose a model that is capable of analysing the content of chats semiautomatically, in order to determine the most important threads that were discussed in CSCL sessions. To do this, it mainly relies on Bakhtin's ideas and Trausan-Matu's polyphonic model. Student dialogues are analyzed in order to determine the moments of convergence and divergence in their points of view, additionally to providing results in statistical tables and graphics. By these results, teachers can evaluate the educational dialogues in order to know whether students concur or not in their points of views. By doing so, this will help students in establishing educational strategies that can lead to an educational collaborative dialogue without stress or selfishness.