byte pair encoding

Fetched 2026-10-26 (cache).

Datamuse reports no synonyms or antonyms for this word.

Noun

1. {{lb|en|computing}} A [[lossless]] data [[compression]] [[algorithm]] that [[iteratively]] replaces the most [[frequent]] pair of [[adjacent]] bytes in a [[sequence]] with a new byte not already present in the [[data]].
2. {{lb|en|natural language processing}} A [[subword]] [[tokenization]] method that iteratively [[merge]]s the most frequent pairs of adjacent characters in a [[corpus]] to form longer and more meaningful [[token]]s, typically until a predefined [[vocabulary]] size is [[reach]]ed.

Source: Wiktionary · CC BY-SA 4.0 (Wiktionary dump) · cached copy, not a rewrite.