Using topic models to bridge coding schemes of differing granularity


While Intelligent Tutoring Systems (ITSs) are often informed by the data extracted from tutoring corpora, coding schemes can be time consuming to implement. Therefore, an automatic classifier may make for quicker classifications. Dialogue from expert tutoring sessions were analyzed using a topic model to investigate how topics mapped on to pre-existing coding schemes of different granularities. These topics were then used to predict the classification of words into moves and modes. Ultimately, it was found that a decision tree algorithm outperformed several other algorithms in this classification task. Improvements to the classifier are discussed.

Publication Title

Educational Data Mining 2010 - 3rd International Conference on Educational Data Mining

This document is currently not available here.