Splintering Nonconcatenative Languages for Better Tokenization
This addresses tokenization challenges for nonconcatenative languages, offering a domain-specific improvement.
The paper tackles the problem of subword tokenization for nonconcatenative languages like Hebrew and Arabic by introducing SPLINTER, a pre-processing step that rearranges text to improve tokenization, resulting in better intrinsic measures and downstream task performance for Hebrew BERT models.
Common subword tokenization algorithms like BPE and UnigramLM assume that text can be split into meaningful units by concatenative measures alone. This is not true for languages such as Hebrew and Arabic, where morphology is encoded in root-template patterns, or Malay and Georgian, where split affixes are common. We present SPLINTER, a pre-processing step which rearranges text into a linear form that better represents such nonconcatenative morphologies, enabling meaningful contiguous segments to be found by the tokenizer. We demonstrate SPLINTER's merit using both intrinsic measures evaluating token vocabularies in Hebrew, Arabic, and Malay; as well as on downstream tasks using BERT-architecture models trained for Hebrew.