Investigating relationships of sentiments, emotions, and performance in professional development K-12 CS teachers

Background and Context: Professional development (PD) programs for K-12 computer science teachers use surveys to measure teachers' knowledge and attitudes while recognizing daily sentiment and emotion changes can be crucial for providing timely teacher support. Objective: We investigate approaches to compute sentiment and emotion scores automatically and identify associations between the scores and teachers' performance. Method: We compute the scores from teachers' assignments using a machine-assisted tool and measure score changes with standard deviation and linear regression slopes. Further, we compare the scores to teachers' performance and post-PD qualitative survey results. Findings: We find significant associations between teachers' sentiment and emotion scores and their performance across demographics. Additionally, we find significant associations that are not captured by post-PD qualitative surveys.