Automatic Analyses of Cohesion and Coherence in Human Tutorial Dialogues During Hypermedia:A Comparison among Mental Model Jumpers

摘要

We analyzed cohesion and coherence in tutorial dialogues from 66 think-aloud transcripts collected from a human tutorial dialogue study which investigated the effect of tutoring on middle and high school students' learning about the circulatory system with hypermedia [1]. Our findings showed that there were significant differences in the tutorial dialogues of Jumpers (i.e., those who showed significant pretest-posttest mental model shifts about the science topic) versus No-jumpers (i.e., those who showed no significant shifts) in the semantic/conceptual similarity, readability scores, incidence scores of causal verbs and causal connectives, and turn length. We argue that the semantic/conceptual similarity of the discourse, causal verbs/causal connectives, and longer turns primarily facilitated the improvement in Jumpers' mental models and deep learning.