Interaction Strategies in Online Learning: Insights from Text Analytics on iMOOC

Learners engaged in large-scale online learning often pose questions in which their peers or instructors can answer using various means of textual interaction topics. This paper assesses the effects of the text interaction strategies in online learning through the lens of the language expectancy theory at three levels: whether to respond to the questions, the identity of the respondents, and the textual interaction topics. Using 112,680 learning records of 610 courses from 71,948 learners crawled from the online learning programming platform iMOOC as the corpus, text mining is used to identify the interaction strategies. Using grounded theory, the textual interaction topics are divided into 2 groups (providing solutions, and encouragement & evaluation for the learners), and sub-divided into 6 topic clusters ("code writing," "operation guidance," "providing references," "encouragement," "normative interpretation," and "opinion exchange"). The responses are classified by text mining. The results of the econometric model suggest that responding to the questions online fosters learning and reduces the dropout rate. The online learner benefits more from peer learning than from the instructors. On the text interaction topics, the topic "providing solutions" is more effective in reducing the learner's dropout rate than the topic "encouragement & evaluation". Further, "code writing" is more effective over "providing references," "encouragement," and "normative interpretation." This study enriches our understanding of the interaction strategies between learners and instructors in iMOOC, and provides a reference for improving the online learning journey and retain learners.