| AutoTutor is a web-based computer tutor architecture that can be used for a variety of content domains. AutoTutor simulates the dialog moves of effective human tutors in conversationally appropriate ways. The Tutoring Research Group has developed two versions of AutoTutor, one for computer literacy and one for conceptual physics. The computer literacy AutoTutor is designed to help students learn basic computer literacy topics covered in a college-level introductory course (e.g., hardware, operating systems, and the Internet). The conceptual physics AutoTutor is designed to help students learn Newtonian physics. AutoTutor works by having a conversation with the learner. AutoTutor includes as an animated agent that acts as a dialog partner with the learner. The animated agent delivers AutoTutor's dialog moves with synthesized speech, intonation, facial expressions, and gestures. Students are encouraged to articulate lengthy answers that exhibit deep reasoning, rather than to recite small bits of shallow knowledge. For some topics, there are graphical displays, animations, and simulations Features of AutoTutor - Animated agent and simulations: AutoTutor is currently a three-dimensional embodied agent who remains on the screen throughout the entire tutoring session. AutoTutor communicates with the learner via synthesized speech, facial expressions, and simple hand gestures. The upcoming version of AutoTutor (due in Fall 2003) will include more dynamic and lifelike means of agent representation and control. Interactive simulations are currently being developed that will be used in tandem with the AutoTutor agent to facilitate learning. The use of interactive simulations coupled with dialog in learning environments will improve communication and visualization of a problem.
- Curriculum Script & Lesson Authoring Tools: Each tutoring topic (or focal question) is organized in a curriculum script. A curriculum script is a structured set of concepts, example problems, dialog moves, and question-answer units. Curriculum scripts also include set of anticipated good and bad answers, corrections for each bad answer, a list of important concepts, and ideal answers to the focal questions. We are currently developing authoring tools that will allow domain experts and novice users to easily build their own curriculum scripts for new content domains.
- Latent Semantic Analysis: Latent semantic analysis (LSA) is a statistical representation of a large body of world knowledge. LSA can be used to determine the conceptual similarity between any two sets of text. When students type in their responses, AutoTutor uses LSA to determine how well the student input matches anticipated good responses in AutoTutor's curriculum script. In addition to assessing the quality of student input, AutoTutor uses LSA to identify misconceptions and to keep track of what has been covered in the tutoring session.
- Speech Act Classifier: For AutoTutor to understand a student, the system's semantic analyses need to be augmented with pragmatic analyses. To capture these aspects of language use, AutoTutor classifies incoming expressions into speech acts. By using a number of advanced computational linguistic techniques, a speech act classifier is able to reliably classify incoming expressions into assertions, directives, meta-cognitive and meta-communicative speech acts, as well as sixteen question categories. The implemented speech act classifier has proven to make the conversation between AutoTutor and the student more natural and more efficient.
- QUEST: Question Answering System: AutoTutor can answer a variety of student questions. The system uses the speech act classifier and relevant queried concepts and returns a relevant paragraph from a textbook. QUEST can deal with a variety of questions, including deep comprehension questions. For instance, depending on what the student asks, it can answer in terms of definitions, examples, causes, effects, and opinions about phenomena and concepts. The capabilities of QUEST are being expanded to provide more fine-tuned answers by taking context and common ground of AutoTutor's conversation into consideration.
You can view our past and current National Science Foundation and Office of Naval Research Grant Proposals to gain a more comprehensive overview of the Autoutor project: View AutoTutor Installation Notes |
