Analysis of Collaborative Writing Processes Using Hidden Markov Models and Semantic Heuristics

In this paper we are interested in discovering collaborative writing patterns in student data collected from a system we designed to support student collaborative writing, and which has been used by over 1,000 students in the past year. A particular functionality that we are investigating is the extraction and display to learners and teachers of the process followed during the course of the writing. We used a heuristic to derive semantic interpretation of specific sequences of raw data and Markov models (MM) to derive the processes. We propose two models, a Heuristic MM and a Hidden MM for analysing student's writing behavior. We also refined the semantic preprocessing by adding the notion of pauses between activities. We illustrate our approach and compare these models using real data from two groups of high and low performance level and highlight the different information they each provide.