The paper shows that, under the assumption that detokenization is compositional, extended tokenization is an inverse homomorphism and thus preserves context-free and regular language structure, while the structure of the tokenizer's unique proper output remains open.
Hopcroft, Rajeev Motwani, and Jeffrey D
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Byte BPE Tokenization as an Inverse string Homomorphism
The paper shows that, under the assumption that detokenization is compositional, extended tokenization is an inverse homomorphism and thus preserves context-free and regular language structure, while the structure of the tokenizer's unique proper output remains open.