Tokens with Meaning: A Hybrid Tokenization Approach for Turkish

TL;DR

This paper presents a linguistically informed hybrid tokenizer for Turkish that improves morphological segmentation and outperforms general-purpose tokenizers on multiple NLP benchmarks.

Contribution

The authors introduce a novel Turkish tokenizer combining morphological, phonological, and subword techniques, with a comprehensive vocabulary and evaluation metrics.

Findings

Tokenizer achieves 90.29% Turkish Token Percentage on TR-MMLU.

Outperforms baselines on Turkish STS Benchmark and MTEB-TR.

Yields strongest accuracy on TurBLiMP under a proxy.

Abstract

Tokenization shapes how language models perceive morphology and meaning in NLP, yet widely used frequency-driven subword tokenizers (e.g., Byte Pair Encoding and WordPiece) can fragment morphologically rich and agglutinative languages in ways that obscure morpheme boundaries. We introduce a linguistically informed hybrid tokenizer for Turkish that combines (i) dictionary-driven morphological segmentation (roots and affixes), (ii) phonological normalization that maps allomorphic variants to shared identifiers, and (iii) a controlled subword fallback for out-of-vocabulary coverage. Concretely, our released Turkish vocabulary contains 22,231 root tokens mapped to 20,000 canonical root identifiers (with leading spaces to mark word boundaries), 72 affix identifiers that cover 177 allomorphic surface forms, and 12,696 subword units; an orthographic case token preserves capitalization without inflating the vocabulary. We evaluate tokenization quality on the TR-MMLU dataset using two linguistic alignment metrics: Turkish Token Percentage (TR~\%), the proportion of produced tokens that correspond to Turkish lexical/morphemic units under our lexical resources, and Pure Token Percentage (Pure~\%), the proportion of tokens aligning with unambiguous root/affix boundaries. The proposed tokenizer reaches 90.29\% TR~\% and 85.80\% Pure~\% on TR-MMLU, substantially exceeding several general-purpose tokenizers. We further validate practical utility with downstream sentence embedding benchmarks under a strict \emph{random initialization} control to isolate tokenizer inductive bias. Across four matched models (TurkishTokenizer, CosmosGPT2, Mursit, and Tabi), TurkishTokenizer outperforms all baselines on the Turkish STS Benchmark and achieves the strongest overall average on MTEB-TR. It also yields the strongest average accuracy on the TurBLiMP under a centroid-based proxy.