Application of sentence-level text analysis: The role of emotion in an experimental learning intervention

Abstract

This registered study aimed at testing the role of emotion in the intervention effect of an experimental intervention study in academic settings. Previous analyses of the National Study of the Learning Mindset (Yeager et al., 2019) showed that in a randomized controlled trial, high school students who were given the growth mindset intervention had, on average higher GPA than did students in the control condition. Previous analyses also showed that school achievement levels moderated the intervention effect. This study applied a sentence-level text analysis strategy to detect participants' attentional focus in five emotional dimensions (valence, arousal, dominance/control, approach-avoidant, and uncertainty) across three writing prompts students wrote during the intervention. Linear mixed models were conducted to test if emotional dimension scores computed using the text analysis predicted a higher intervention effect (i.e., higher post-intervention GPA given pre-intervention GPA). The moderating role of school achievement levels was also examined. The results of this study have implications on the possibility of applying text analysis strategies on open-ended questions in interventions or experimental studies to examine the role of the emotion-attentional focus of participants during intervention or experimental studies on the intervention or experimental outcomes, especially those that are conducted in academic settings.

Activism through fandom for the Black Lives Matter movement.

“…Furthermore, sentiments (a continuous variable from negative to positive valence) were analyzed using sentence-level text analysis R package, sentimentr. sentimentr was previously found to be reliable and valid in multiple studies (e.g., Li, 2022;Naldi, 2019). Positive emotions, negative emotions, and sentiments were then correlated with the codes and themes identified to describe the emotions/sentiments reflected in each coding category.…”

Overview of the Impact of Collaborative Online International Learning on Learners

“…Emotions can be expressed through verbal or nonverbal communication [10] and affect learning outcomes. Students with positive emotions can show higher achievement than those with less positive emotions [11]. However, COIL students receive insufficient guidance to regulate their emotional engagement, and therefore some students might refrain from further engagement because of the emotional fluctuation during COIL, especially for the tasks after ice-breaker activities.…”

TERMS: textual emotion recognition in multidimensional space

“…Cheng et al proposed a Bi-directional Long Short-Term Memory (BiLSTM) model that identifies and forecasts the sentiment information in terms of VA-values and integrated it into a deep learning model to optimise Government social management [ 51 ]. Another recent experimental work aimed at testing the role of five emotions (valence, arousal, dominance, approach-avoidant, and uncertainty) on the intervention effect of the Learning Mindset study [ 52 ]. The SemEval-2018 Task 1: Affect in Tweets challenge asked for the prediction of intensities (arousal) and valence from a stream of texts in terms of regression and ordinal classification [ 37 , 53 ].…”