Using LSA in AutoTutor: Learning Through Mixed-Initiative Dialogue in Natural Language

Abstract

AutoTutor is a computer tutor that holds conversations with students in natural language (Graesser, Hu, and McNamara, 2005; Graesser, Lu, et al., 2004; Graesser, Person, Harter, and the Tutoring Research Group, 2001; Graesser, VanLehn, Rose, Jordan, and Harter, 2001; Graesser, K. Wiemer-Hastings, P. Wiemer-Hastings, Kreuz, and Harter, 1999). AutoTutor simulates the discourse patterns of human tutors and a number of ideal tutoring strategies. It presents a series of challenging problems (or questions) from a curriculum script and engages in collaborative, mixed initiative dialog while constructing answers. AutoTutor speaks the content of its turns through an animated conversational agent with a speech engine; it was designed to be a good conversational partner that comprehends, speaks, points, and displays emotions, all in a coordinated fashion. For some topics, there are graphical displays, animations of causal mechanisms, or interactive simulation environments (Graesser, Chipman, Haynes, and Olney, 2005). So far, AutoTutor has been developed and tested for topics in Newtonian physics (VanLehn et al., in press) and computer literacy (Graesser, Lu, et al., 2004), showing impressive learning gains compared to pretest measures and suitable control conditions. One notable characteristic of AutoTutor, from the standpoint of the present edited volume, is that latent semantic analysis (LSA) was adopted as its primary representation of world knowledge.

Publication Title

Handbook of Latent Semantic Analysis

Share

COinS