Comments of Journalism Mentors on News Stories: Classification and Epistemic Status of Mentor Contributions

We identified the speech act categories and clusters of discourse comments of journalism mentors who interact with students editing news stories. Two important speech act categories are evaluations and suggestions. Latent semantic analysis and principal components analyses helped us discover clusters of comments involving evaluations and suggestions. The comments of mentors were also significantly aligned with epistemic frame elements that motivate the comments at a deeper level of discourse and pedagogy. Such alignments were validated by logistic regression analyses on a sample of handcoded judgments of the frame elements. There was some modest transfer from a journalism practicum corpus to a game corpus. These analyses provide an important first step in building a virtual AutoMentor for multiparty epistemic games.