<![CDATA[The training program for Senior and Vocational High School teachers of the Coding and Artificial Intelligence (KKA) subject serves as a strategic effort to strengthen human resource capacity in response to digital transformation. Teachers are expected to understand the KKA learning outcomes and be able to design and implement technology-based learning in their respective schools. This study employs a Large Language Model (LLM)-based categorization technique to analyse the reflective content of SMA/SMK teachers in Malang City after attending the KKA training, aiming to identify their learning achievement levels and how they interpret their learning experiences. The dataset, consisting of presentation files in PDF and PPTX formats, was text-extracted, processed by the LLM, categorized into ten main themes in JSON format, and visualized as a mind map using Mermaid scripts. The findings demonstrate that the LLM effectively performs consistent semantic categorization on unstructured reflective data and produces visual representations that facilitate the interpretation of learning achievements. Content analysis of teacher reflections reveals strong motivation to enhance digital and pedagogical competencies, successful implementation of project-based learning, and emerging challenges related to student scepticism toward AI adoption in art and design education. Furthermore, the comparison between a general-purpose LLM (GPT-OSS) and coder-type LLMs (such as GLM-4.6:Coder and Qwen3-Coder) indicates comparable accuracy in generating JSON structures and Mermaid scripts. However, the non-coder LLM exhibits greater stability in maintaining contextual coherence and processing speed. These findings highlight the potential of LLMs in analysing reflective educational data and underscore the need for pedagogical adaptation and continued digital competence development among teachers toward AI-driven educational independence.]]>
Copyrights © 2025
