Development of Physics Text Corpora for Latent Semantic Analysis
Donald R. Franceshetti, Ashish Karnavat, Johanna Marineau, Genna L. McCallie, Brent A. Olde, Blair L. Terry, Arthur C. Graesser · eScholarship (California Digital Library) · 2001
Student responses to qualitative physics questions were analyzed with latent semantic analysis (LSA), using different text corpora.Physics potentially has a number of distinctive characteristics that are not encountered in many other knowledge domains.Physics texts exist at a variety of levels and typically involve an integrated presentation of text, figures and equations.We explore the adequacy of several text corpora and report results on vector lengths and correlations between key terms in elementary mechanics.The results suggest that a carefully constructed smaller corpus may provide a more accurate representation of fundamental physical concepts than a much larger one.