This meta-analysis explores the correlation between self-assessment (SA) and language performance. Sixty-seven studies with 97 independent samples involving more than 68,500 participants were included in our analysis. It was found that the overall correlation between SA and language performance was .466 ( p < .01). Moderator analysis was performed to evaluate the moderating effects of a number of variables that are believed to play important roles in SA. Six moderators were found to have significant moderating effects, including SA criteria type, presence and form of SA criteria, SA instruments, training, total number of items in the SA instrument, and reliability of SA instruments. However, no significant moderating effect was identified for external measures, language skills, order of assessment, reliability of external measures, and number of Likert-scale levels. These findings offer a number of implications for improving the SA–performance correlation.

The evaluation of teaching activities plays an integral part in measuring teachers’ abilities and improving teaching qualities. Among of which, Student Evaluation of Teaching (SET) serves as the key approach to assessing teaching activities. In the field of sports teaching in China, SET has been implemented nationwide, however, very few studies have touched upon the aspect of measuring sports teaching activities in higher education from the perspective of SET. Among rare studies, attention has been mainly directed to aspects of structured data (e.g., rating scales), which are easy to quantify, rather than descriptive evaluation comments/reviews that came from students (e.g., texts). Natural Language Processing (NLP) is deemed as an effective tool to deal with unstructured data and may explore more valuable diagnostic information through a multidimensional view (i.e., data mining), thus improving sports teaching and promoting learning in the long run. This research collected SET data from several universities. The NLP approach and Latent Dirichlet Allocation (LDA) topic model were used to analyze the SET corpora for feature extraction. The results were then used to identify traits that may be considered for measuring sports teaching activities. Besides, a user dictionary was constructed through thematic coding and was then further applied to investigate the emotional polarity and subjectivity of SET.