Automated Sentiment Analysis of Peer Feedback Patterns in Collaborative Writing Environments for Higher Education

Abstract

With the rise in the use of collaborative writing in higher education, the importance of peer feedback in developing learners' writing and critical analytical skills has become more significant. Determining the quality and tone of feedback has its own challenges due to subjectivity and large-scale issues. In this regard, this study aims to introduce an automated framework for conducting basic analysis of peer feedback in collaborative writing. 1150 peer feedback comments from collaborative writing in undergraduate English classes were examined using a hybrid Natural Language processing framework that integrates lexicon and machine learning. Feedback was analysed by dividing comments into positive, negative, and neutral sentiments using automated text preprocessing and feature extraction based on domain-specific symbols. The findings illustrate that the majority of comments were positive at 64.2%, and only 15.3% of comments were negative. The analysis framework registered 90.3% correct feedback sentiment assessment, 89.7% correct predictions of positive feedback, 90.3% of correct assessments of negative feedback, and 89.6% correct assessments of neutral feedback, which indicates feedback was analysed reliably. Further, positive feedback correlated with improvement of writing drafts to a fair extent (r = 0.71). Evidently, analysis of feedback interaction and quality showed the potential benefits of sentiment analysis in Peer Feedback. The study indicates that the application of basic sentiment analysis in a collaborative writing platform (for higher education), could significantly strengthen the monitoring of feedback in writing, enhance the collaborative writing, and provide the instructors with tools to handle large classes. The study represents the intersection of modern English education and Artificial Intelligence, and analyses the feedback of peer writing through a unique and scalable approach.