tailieunhanh - ARTIFICIAL_INTELLIGENCE_IN_EDUCATION 2

(BQ) Part 2 book "Artificial intelligence in education" has contents: Linguistics and language technology, tutorial dialogues; collaboration, course based experimentation; authoring tools and ontologies, data mining, young researchers’ track abstracts, workshop summaries, workshop summaries, interactive events summaries,. and other contents. | Linguistics and Language Technology, Tutorial Dialogues This page intentionally left blank Artificial Intelligence in Education R. Luckin et al. (Eds.) IOS Press, 2007 © 2007 The authors and IOS Press. All rights reserved. 341 Analyzing the Coherence and Cohesion in Human Tutorial Dialogues when Learning with Hypermedia Roger AZEVEDO and Moongee JEON Dept. of Psychology, Institute for Intelligent Systems, University of Memphis 3693 Norriswood Avenue, Memphis, TN, 38152, USA {razevedo@; mjeon1@} Abstract. We examined the coherence and cohesion from 38 think-aloud transcriptions from a human tutorial dialogue study examining the role of tutoring on college students’ learning about the circulatory system with hypermedia. The corpus we examined had a total of 800 pages. We used Coh-Metrix, a web-based tool designed to examine text and discourse, to evaluate the coherence and cohesion of the text produced between a human tutor and low-domain knowledge college students during 38 tutoring sessions. Our findings demonstrated that there were significant differences in the tutorial dialogues of the high-jump students (., those who showed conceptual gains in their pretest-posttest mental models of the circulatory system) versus the medium-jumpers or no-jumpers in the semantic/conceptual overlap, the negative additive connectives incidence scores, the number of turns, and the average words per sentence. However, the tutorial dialogues of the high jump students substantially shared the syntactic and linguistic similarities with the tutorial dialogues of the medium or no jump students in the standard readability formulas, the coreferential cohesion (argument and stem overlaps), and the incidence scores of all the connectives. We argue that the semantic/conceptual overlap of the tutorial dialogues primarily promoted the high “jumps” or improvement in students’ mental model and deep learning while interacting with a tutor. Our findings have .