Character-based language model approach. They should have predicted the next syllable/character, at least in some poets. I am quite confidence that the extracted equivalent rules/automata/templates from many models (for linguistically zcs#op23$ verification) are nonsense.
Another promising way might be (unsupervised) sentencepiece/BPE. I am quite surprised to see what Thai word segmentation methods are for?
Character-based language model approach. They should have predicted the next syllable/character, at least in some poets. I am quite confidence that the extracted equivalent rules/automata/templates from many models (for linguistically zcs#op23$ verification) are nonsense.
Another promising way might be (unsupervised) sentencepiece/BPE. I am quite surprised to see what Thai word segmentation methods are for?