Understanding Pre-service Teachers' Artificial Intelligence Perceptions Through Text-Mining: A Mixed-Method Analysis of Language Patterns, Themes, and Sentiment

As artificial intelligence rapidly transforms educational landscapes, pre-service teachers' perceptions and preparedness become pivotal for successful technology integration. While existing research has examined broad attitudes toward AI in education, a significant gap exists in understanding the nuanced linguistic, thematic, and emotional dimensions of how future educators conceptualize AI. This study employed textmining techniques, using topic modeling and sentiment analysis to examine pre-service teachers' AI perceptions in Hong Kong. Participants provided five adjectives in Hong Kong Chinese to describe AI, yielding 258 responses. Data underwent frequency analysis, BERTopic modeling for thematic clustering, and AIassisted sentiment classification. Frequency analysis revealed 'convenient' (13.2 %) and 'expedient' (8.1 %) as dominant descriptors. Topic modeling identified 15 themes, with 'Usability and Every day Convenience' (14.75%) and 'Technology's Societal Impact and Debate' (14.34 %) as the most significant clusters. Sentiment analysis showed 73.3 % positive responses, with all 10 participants reporting previous AI experiences classified as having positive perceptions, while none of the neutral/negative participants reported previous AI experiences. The findings reveal predominantly positive but potentially surface-level AI perceptions among pre-service teachers, with a strong relationship between experience and positive sentiment. These insights inform evidence-based approaches to teacher preparation programs, emphasizing hands-on AI experiences and deeper pedagogical understanding beyond convenience-focused perceptions.