Analysing EFL Teachers’ Perceptions of AI’s Role in Academic Integrity and Pedagogy with BERT-LSTM

The increasing adoption of artificial intelligence in education raises worries about its effects on academic integrity and teaching methods. Conventional sentiment analysis approaches improperly capture complex perceptions, requiring the use of more sophisticated methodologies. This study presents a sentiment analysis method based on BERT-LSTM to investigate EFL instructors' perspectives on the use of AI in education. The suggested model adeptly integrates BERT’s contextual word embeddings with LSTM’s sequence modelling, resulting in enhanced classification performance. Experimental findings indicate a notable enhancement compared to conventional models, with accuracy at 96.2%, precision at 95.8%, recall at 96.5%, F1-score at 96.1%, and ROC-AUC at 98.3%. The comparative comparison of baseline models, including SVM, Random Forest, and CNN-LSTM, validates the superiority of our methodology. The study's results offer significant insights for educators, policymakers, and AI developers, facilitating ethical AI integration in education. This study connects AI-based sentiment analysis with educational decision-making, enhancing confidence in AI-supported learning scenarios.