VerChol -- Grammar-First Tokenization for Agglutinative Languages

TL;DR

VerChol introduces a grammar-first tokenization approach tailored for agglutinative languages, addressing the limitations of BPE by respecting morpheme boundaries and improving tokenization accuracy.

Contribution

It proposes a novel grammar-based tokenization method specifically designed for agglutinative languages, enhancing morphological preservation over traditional statistical methods.

Findings

Reduces token count inflation in agglutinative languages

Improves morpheme boundary detection accuracy

Outperforms BPE-based tokenizers on multiple languages

Abstract

Tokenization is the foundational step in all large language model (LLM) pipelines, yet the dominant approach Byte Pair Encoding (BPE) and its variants is inherently script agnostic and optimized for English like morphology. For agglutinative languages a typological class encompassing the Dravidian family (Tamil, Kannada, Telugu, Malayalam), Turkic languages (Turkish, Azerbaijani, Uzbek), Uralic languages (Finnish, Hungarian, Estonian), Korean, Japanese, Swahili, Basque, and others, a single word may encode root, tense, aspect, person, number, gender agreement, case, and postpositions into one orthographic unit. Statistical tokenizers fragment these words into byte pair chunks that sever morpheme boundaries and inflate token counts.