Contribution: An AI model for speech emotion recognition (SER) in the educational domain to analyze the correlation between students' emotions, discussed topics in teams, and academic performance. Background: Research suggests that positive emotions are associated with better academic performance. On the other hand, negative emotions have a detrimental impact on academic achievement. This highlights the importance of taking into account the emotional states of the students to promote a supportive learning environment and improve their motivation and engagement. This line of research allows the development of tools that allow educators to address students' emotional needs and provide timely support and interventions. Intended Outcome: This work analyzes students' conversations and their expressed emotions as they work on class activities in teams and investigates if their conversations are course-related or not by applying topic extraction to the conversations. Furthermore, a comprehensive analysis is conducted to identify the correlation between emotions expressed by students and the discussed topics with their performance in the course in terms of their grades. Application Design: The student's performance is formatively evaluated, taking into account a combination of their scores in various components. The core of the developed model comprises a speech transcriber module, an emotion analysis module, and a topic extraction module. The outputs of all these modules are processed to identify the correlations. Findings: The findings show a strong positive correlation between the expressed emotions of "relief" and "satisfaction" with students' grades and a strong negative correlation between "frustration" and grades. Data also shows a strong positive correlation between course-related topics discussed in teams and grades and a strong negative correlation between noncourse-related topics and grades.