Interesting paper (13 page PDF) discussing the use of machine learning to classify the thousands of diiscussion posts in a popular MOOC to help sort through the clutter and find those that are on topic. The purpose of this work isn't so much to hlp students find posts as it is to enable an analysis of the relation between using the discussion board and liklihood of success in the course. The authors are a bit quick to assign agency (eg., saying such-and-such a usage pattern leads to success in the course) but overall this is a clear and well-written article that describes a process that could be replicated by others (and this should be the standard for published reserach articles), though the link to the full dataset (section 3.3) is missing, and the tools used (openNLP, quanteda, weka) are mentioned only in passing. I'm not sure how long this paper will be available so grab it while you can. I have a copy if access closes down.
Today: 2 Total: 22 [Direct link] [Share]
