Cognitive engagement is a crucial factor that shapes successful learning outcomes, but our understanding of the factors that influence such engagement in the smart classroom context remains limited. This study aims to narrow this research gap by exploring the relationships among college students’ perceptions of the smart learning environment, perceived usefulness of mobile technology, achievement emotions, and cognitive engagement. A total of 1293 college students completed an online questionnaire survey, and 1076 valid responses were received. Structural equation modeling was used to analyze the interrelationships among these factors. The results revealed that students’ perceptions of the smart classroom environment and perceived usefulness of mobile technology as well as two achievement emotions (pride and anxiety) significantly impact cognitive engagement. Both pride and anxiety act as mediators in the relationships among perceptions of smart classroom environments, the perceived usefulness of mobile technology, and cognitive engagement, in which context the mediating effect of pride is stronger than that of anxiety. These findings have practical implications for instructors, who should focus on implementing strategies that promote positive achievement emotions when students use mobile technology in smart classrooms. Additionally, these findings can inform the design and construction of smart classroom environments. Moreover, our study has limitations due to reliance on online data collection and self-reported data, which may introduce biases and measurement errors. Future research should incorporate multimodal data and advanced technologies for a comprehensive assessment to better understand students’ engagement in smart learning environments, while also considering individual factors and the educational context to enhance the effectiveness of mobile technology in supporting students’ emotions and achievement.

Teacher professional development (TPD) is increasingly mediated by online platforms, yet emotion analysis in this context remains underdeveloped because teachers’ professional discourse is often reflective, evaluative, and shaped by professional norms. To address this challenge, this study proposes a fine-grained, low-intrusion affective-appraisal analysis framework for TPD-oriented online professional learning that integrates textual evidence with platform interaction logs. The framework retains pleasure, arousal, and dominance from the pleasure–arousal–dominance (PAD) model and introduces utility as an appraisal-related dimension, capturing teachers’ perceived usefulness, value judgment, and professional learning gain. Methodologically, it combines textual representations based on Bidirectional Encoder Representations from Transformers (BERT), intra-week long short-term memory (LSTM) aggregation, interpretable behavioral-log features, and feature-level fusion. Data were collected from an authentic TPD-oriented online course involving 107 pre-service teachers, yielding 1276 teacher-week samples from 4300 texts and 264,028 interaction records. Results show that intra-week sequential modeling improves the macro-averaged F1 score (Macro-F1) over both the term frequency–inverse document frequency plus support vector machine (TF-IDF+SVM) baseline and BERT-based weekly text concatenation, with statistically significant gains over the non-sequential BERT-concat model across all four dimensions. Adding interaction logs improves accuracy across all dimensions and provides complementary process-based evidence, especially for arousal and utility. By linking a four-dimensional affective-appraisal framework with text-log fusion, this study offers a scalable and context-sensitive approach to affective-appraisal analytics in pre-service teacher professional learning.