Maximal Phrases Based Analysis for Prototyping Online Discussion Forums Postings

Chat texts produced in an educational environment are categorized and rated for the purpose of positioning (or placement) of the learner with respect to a learning program (appropriate courses, textbooks, etc). The difficulty lies in the fact that the texts are short and informal. A standard LSA/vector-space model is therefore combined with techniques appropriate for short texts. The approach uses phrases rather than words in the term-document matrix, and for determining prototypical documents of each category, a nonparametric permutation test is used.