Investigating Accuracy of Self-assessment of English Speaking Proficiency Levels by Engineering Students Based on Correlation and Sentiment Analysis

Abstract

The study investigated the accuracy of self-assessment (SA) and possible reasons for inaccuracies using correlation analysis and sentiment analysis based on machine learning. Results show that there was a moderate correlation between self-assessment and teacher assessment, and a general trend of underestimation in student self-assessment, particularly for high-performing and medium-performing students. Results also show that students' lack of confidence in English language learning, confusion about their own speaking abilities, and lack of a clear understanding of the rating scale are likely to be the reasons for underestimation. Though the accuracy of self-assessment was not validated in this study, it could be argued that SA can still be used as a formative assessment tool to trigger active learning and reflection on own abilities and improve assessment literacy and academic performance through using relevant rating scales or standards.