The Beetle II tutorial dialogue system is designed to accept unrestricted language input and to support experimentation with different approaches to tutoring. Previous research in tutoring demonstrates that encouraging students to produce explanations and giving them detailed feedback is important for effective learning. However, understanding natural language input is a major challenge. Many existing tutorial dialogue systems severely restrict the language input they can accept to make interaction possible, or else limit the type and variety of feedback that can be given to the student.

The Beetle II tutor asks students to explain their reasoning and accepts complex, sentence-long answers to such open-ended questions. At present, it uses a deep parser and interpreter to produce detailed analyses of the student's input. A tutoring module makes decisions about which tutoring strategy to apply, choosing from up to seven different possible tactics depending on the problem. Tutoring decisions take into account features of the interaction such as the number and type of previous student mistakes. Finally, a generation module is used to automatically generate appropriate tutoring feedback. This architecture has proven effective for a moderate-size domain (vocabulary of about 2000 words, and over 200 different questions that the system asks during a 3-hour course). Currently, we are looking at combining our approach with shallow interpretation techniques in order to improve system robustness without sacrificing the benefits of the detailed semantic analysis produced by the system.

In the past, we also used a similar architecture in the LEActiveMath tutoring system, which supported students in learning symbolic differentiation, and adapted to both the student's performance, and to the signs of uncertainty in the student's input.

Beetle-Grow, our new web-enabled version of the Beetle II system, collects but does not currently parse student input. Its design is intended to make interaction with the system more intuitive and will facilitate evaluation of the system with increased numbers of students in the future.