Latent Semantic Analysis

Five years ago, it seemed inconceivable to me that I would someday use Latent Semantic Analysis (LSA) to assess online discourse data. I recall vehemently arguing against LSA in a certain online graduate course at OISE, in which Teplovs (2001) was one of the assigned readings. I came up with rather elaborate examples in which I felt LSA would fail.

My thoughts on LSA took a radical turn last spring. That’s when Chris showed me what the Knowledge Space Visualizer could do to pinpoint activity of interest in my thesis data. This had immense appeal for me, as I had spent weeks if not months laboring on qualitative content analysis in NVivo, and the KSV could identify networks of ideas–and social networks–worth examining in far less time. This led to work on our CSCL 2009 poster. We are now preparing to present a more complete analysis for the Pinpointing Pivotal Moments in Collaboration Workshop at Alpine Rendez-Vous 2009 in Garmisch-Partenkirchen.


Leave a Reply

Your email address will not be published. Required fields are marked *