<![CDATA[This research addresses automated news classification for the Tetun language, Timor-Leste's national language, which remains underrepresented in NLP research. We constructed a machine learning framework to categorize Tetun news headlines from Tatoli.tl and DiliPostNews.com. Our contributions encompass: a specialized Tetun stopwords collection (85 words); a multi-source dataset of 37 articles across 5 categories; and comparative evaluation of four algorithms. Our optimal model attained 75% accuracy, exceeding the majority class baseline (70.27%) and random guessing (14.29%). Analysis revealed language mixing (51.4% Tetun, 32.4% mixed, 16.2% Portuguese). This study provides a proof-of-concept foundational groundwork for Tetun NLP applications.]]>
Copyrights © 2026
