AIC Seminar Series

Intelligent Tutoring Systems (ITSs) with natural language input have been developed at a fast pace recently with the goal of optimizing students’ learning outcomes. Talking is an important aspect of deep understanding of complex topics such as science topics. Possible sources of natural language input in ITSs could be a dialog between the virtual tutor and the tutee, sentences or paragraphs generated by students as part of specific activities they must perform, e.g. paraphrasing a scientific statement, or full-length essays. All these natural language inputs must be assessed by the tutoring system in order to provide adequate feedback to students. This talk will present solutions to automatically assess natural language student inputs of various lengths in ITSs. The focus is on Metatutor, an adaptive hypermedia learning environment that provides training to students on the use of meta-cognitive strategies while learning about complex science topics. In particular, two activities in MetaTutor that require natural language input will be discussed: planning and prior knowledge activation (PKA). During planning, students must specify in their own words several learning subgoals related to the overall learning goal provided by the system. The subgoals must be provided at the optimum level of specificity, neither too specific nor too general. Through a dialog, MetaTutor guides students towards optimum articulation of the subgoals. A taxonomy-driven dialogue manager has been proposed that provides results that correlate highly with experts. During prior knowledge activation, students are prompted to type in a paragraph describing their prior knowledge related to the overall learning goal. Based on the PKA paragraph, the system automatically assesses students’ initial level of understanding of the topic, i.e. students’ mental model. The task of student mental model detection in MetaTutor is modeled as a text classification task, where several machine learning algorithms are experimented with, in order to automatically induce classifiers.

Mihai Lintean received his Bachelors and Masters degree in Computer Science from Babes Bolyai University of Cluj-Napoca, Romania. He is currently a Ph.D. student in the Computer Science Department and a junior researcher in the Institute for Intelligent Systems (IIS) at University of Memphis. Mihai’s primary research interests are in Natural Language Processing, in particular on problems of assessing semantic similarity between texts. Other interests include Machine Learning, Data Mining and Intelligent Systems in general.

Please arrive at least 10 minutes early in order to sign in and be escorted to the conference room. SRI is located at 333 Ravenswood Avenue in Menlo Park. Visitors may park in the visitors lot in front of Building E, and should follow the instructions by the lobby phone to be escorted to the meeting room. Detailed directions to SRI, as well as maps, are available from the Visiting AIC web page.

©2013 SRI International 333 Ravenswood Avenue, Menlo Park, CA 94025-3493
SRI International is an independent, nonprofit corporation. Privacy policy