How do different tokenizers perform on downstream tasks in scriptio continua languages?: A case study in Japanese
This study addresses the problem of tokenizer selection for researchers and practitioners working with scriptio continua languages, providing empirical guidance, though it is incremental as it builds on existing tokenizer methods.
This paper investigated how different tokenizers affect the performance of pretrained language models in scriptio continua languages like Japanese, finding that each downstream task has a different optimal morphological analyzer and that Byte-Pair-Encoding or Unigram tokenizers outperform WordPiece across tasks.
This paper investigates the effect of tokenizers on the downstream performance of pretrained language models (PLMs) in scriptio continua languages where no explicit spaces exist between words, using Japanese as a case study. The tokenizer for such languages often consists of a morphological analyzer and a subword tokenizer, requiring us to conduct a comprehensive study of all possible pairs. However, previous studies lack this comprehensiveness. We therefore train extensive sets of tokenizers, build a PLM using each, and measure the downstream performance on a wide range of tasks. Our results demonstrate that each downstream task has a different optimal morphological analyzer, and that it is better to use Byte-Pair-Encoding or Unigram rather than WordPiece as a subword tokenizer, regardless of the type of task.