Carolyn Penstein Rosé
Language Technologies Institute
Carnegie Mellon University
|Type of Study:||task|
In accordance with TalkBank rules, any use of data from this corpus must be accompanied by at least one of the above references.
This corpus was collected as part of the Computational Model of Tutorial Dialogue project, under the supervision of Johanna Moore, under funding provided by the Office of Naval Research, Cognitive and Neural Sciences Division, Grants N00014-91-J-1691 and N00014-93-I-0812. The purpose of this study was to collect examples of students interacting with a human tutor while working through a computer based basic electricity and electronics curriculum in order to evaluate the relative effectiveness of two alternative tutoring strategies. The first tutoring strategy is a Socratic style strategy in which the tutor attempts to encourage the student to construct knowledge for himself by leading him through a Socratic style directed line of reasoning. In the second strategy, a more Didactic style, the tutor is more forthcoming with explanations up front but then encourages the student to actively engage in the learning process by making an inference, rephrasing, or applying the knowledge from the explanation.
We started with an available BE&E course developed in the VIVIDS authoring environment at NPRDC. This course included four lessons covering Current, Voltage (both AC and DC), Resistance and Power. It also included four labs allowing the students to make basic measurements using a multimeter. To this original curriculum, we added two problem solving labs covering Ohm's Law and the Power Law. 44 students participated in our experiment. Each student's participation spanned over two sessions of 2 - 2.5 hours and included a pre and post-test in addition to the computer based curriculum. The student interacted with the on-line BE&E course as well as with a human tutor through a chat interface. Tutor and student were in the same room but separated by a partition. The student's video signal was split so that the tutor was able to monitor the student's progress through the curriculum. Beginning with a small pilot study, we collected data from 8 students. Based on our observations from this small study of which parts of the curriculum were problematic for students, we focused our pre-post test used in the final experiment. In this final experiment, we collected data from 36 students, 20 of which we used for a rule gain analysis in order to draw some preliminary conclusions about the relative effectiveness of our two alternative tutoring strategies. For each of the 36 students who participated in the final experiment, this corpus includes a log file from each of the two sessions they participated in, a calculations file that records all of the operations they performed using an on-line calculator during the lessons, and a notes file that lists the condition the student was assigned to, Math and Verbal SAT scores when available, dates of the two sessions, pre and post test score, and time on task for each lesson and lab. No special annotations are included in the log files apart from a time stamp for each contribution to the dialogue from student and tutor. A coding scheme is currently being developed at the University of Edinburgh.