Scaling BERT Models for Turkish Automatic Punctuation and Capitalization Correction

The paper evaluates BERT-based models for Turkish punctuation and capitalization correction, focusing on five sizes (Tiny, Mini, Small, Medium, Base). These models, optimized for Turkish challenges, aim to balance performance with minimal computational overhead. The study compares models' precision, recall, and F1 score, with the Base model showing the highest correction precision. It offers guidance for selecting the appropriate model based on user needs and computational resources, enhancing Turkish text quality in real-world applications.