The Art of Breaking Words: Rethinking Multilingual Tokenizer Design

• URL: http://arxiv.org/abs/2508.06533v1
• Date: Sun, 03 Aug 2025 15:31:10 GMT
• Title: The Art of Breaking Words: Rethinking Multilingual Tokenizer Design
• Authors: Aamod Thakur, Ajay Nagpal, Atharva Savarkar, Kundeshwar Pundalik, Siddhesh Dosi, Piyush Sawarkar, Viraj Thakur, Rohit Saluja, Maunendra Sankar Desarkar, Ganesh Ramakrishnan,
• Abstract summary: Existing tokenizers exhibit high token-to-word ratios, inefficient use of context length, and slower inference.<n>We present a systematic study that links vocabulary size, pre-tokenization rules, and training-corpus composition to both token-to-word efficiency and model quality.<n>Our tokenizer achieves more than 40% improvement on average token-to-word ratio against stateof-the-art multilingual Indic models.
• Score: 21.9940001977516
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While model architecture and training objectives are well-studied, tokenization, particularly in multilingual contexts, remains a relatively neglected aspect of Large Language Model (LLM) development. Existing tokenizers often exhibit high token-to-word ratios, inefficient use of context length, and slower inference. We present a systematic study that links vocabulary size, pre-tokenization rules, and training-corpus composition to both token-to-word efficiency and model quality. To ground our analysis in a linguistically diverse context, we conduct extensive experiments on Indic scripts, which present unique challenges due to their high script diversity and orthographic complexity. Drawing on the insights from these analyses, we propose a novel algorithm for data composition that balances multilingual data for tokenizer training. Our observations on pretokenization strategies significantly improve model performance, and our data composition algorithm reduces the average token-to-word ratio by approximately 6% with respect to the conventional data randomization approach. Our tokenizer achieves more than 40% improvement on average token-to-word ratio against stateof-the-art multilingual Indic models. This improvement yields measurable gains in both model performance and inference speed. This highlights tokenization alongside architecture and training objectives as a critical lever for building efficient, scalable multilingual LLMs

Related papers

• Asm2SrcEval: Evaluating Large Language Models for Assembly-to-Source Code Translation [4.45354703148321]
  Assembly-to-source code translation is a critical task in reverse engineering, cybersecurity, and software maintenance.<n>We present the first comprehensive evaluation of five state-of-the-art large language models on assembly-to-source translation.
  arXiv  Detail & Related papers  (2025-11-28T12:40:30Z)
• Tokenization Disparities as Infrastructure Bias: How Subword Systems Create Inequities in LLM Access and Efficiency [6.943451388015595]
  Tokenization disparities pose a significant barrier to achieving equitable access to artificial intelligence.<n>This study conducts a large-scale cross-linguistic evaluation of tokenization efficiency in over 200 languages.
  arXiv  Detail & Related papers  (2025-10-14T11:14:38Z)
• Tokenization Standards for Linguistic Integrity: Turkish as a Benchmark [0.29687381456163997]
  Tokenization is a fundamental preprocessing step in NLP, directly impacting large language models' ability to capture syntactic, morphosyntactic, and semantic structures.<n>This paper introduces a novel framework for evaluating tokenization strategies, addressing challenges in morphologically rich and low-resource languages.
  arXiv  Detail & Related papers  (2025-02-10T21:47:49Z)
• Evaluating Large Language Models Using Contrast Sets: An Experimental Approach [0.0]
  We introduce an innovative technique for generating a contrast set for the Stanford Natural Language Inference dataset. Our strategy involves the automated substitution of verbs, adverbs, and adjectives with their synonyms to preserve the original meaning of sentences. This method aims to assess whether a model's performance is based on genuine language comprehension or simply on pattern recognition.
  arXiv  Detail & Related papers  (2024-04-02T02:03:28Z)
• Split and Rephrase with Large Language Models [2.499907423888049]
  Split and Rephrase (SPRP) task consists in splitting complex sentences into a sequence of shorter grammatical sentences. We evaluate large language models on the task, showing that they can provide large improvements over the state of the art on the main metrics.
  arXiv  Detail & Related papers  (2023-12-18T10:16:37Z)
• Tokenizer Choice For LLM Training: Negligible or Crucial? [30.33170936148845]
  We study the influence of tokenizer choice on Large Language Models (LLMs) downstream performance by training 24 mono- and multilingual LLMs. We find that the tokenizer choice can significantly impact the model's downstream performance and training costs. We show that multilingual tokenizers trained on the five most frequent European languages require vocabulary size increases of factor three in comparison to English.
  arXiv  Detail & Related papers  (2023-10-12T22:44:19Z)
• CompoundPiece: Evaluating and Improving Decompounding Performance of Language Models [77.45934004406283]
  We systematically study decompounding, the task of splitting compound words into their constituents. We introduce a dataset of 255k compound and non-compound words across 56 diverse languages obtained from Wiktionary. We introduce a novel methodology to train dedicated models for decompounding.
  arXiv  Detail & Related papers  (2023-05-23T16:32:27Z)
• A Multi-dimensional Evaluation of Tokenizer-free Multilingual Pretrained Models [87.7086269902562]
  We show that subword-based models might still be the most practical choice in many settings. We encourage future work in tokenizer-free methods to consider these factors when designing and evaluating new models.
  arXiv  Detail & Related papers  (2022-10-13T15:47:09Z)
• An Empirical Investigation of Commonsense Self-Supervision with Knowledge Graphs [67.23285413610243]
  Self-supervision based on the information extracted from large knowledge graphs has been shown to improve the generalization of language models. We study the effect of knowledge sampling strategies and sizes that can be used to generate synthetic data for adapting language models.
  arXiv  Detail & Related papers  (2022-05-21T19:49:04Z)
• Improving Classifier Training Efficiency for Automatic Cyberbullying Detection with Feature Density [58.64907136562178]
  We study the effectiveness of Feature Density (FD) using different linguistically-backed feature preprocessing methods. We hypothesise that estimating dataset complexity allows for the reduction of the number of required experiments. The difference in linguistic complexity of datasets allows us to additionally discuss the efficacy of linguistically-backed word preprocessing.
  arXiv  Detail & Related papers  (2021-11-02T15:48:28Z)
• Distributionally Robust Multilingual Machine Translation [94.51866646879337]
  We propose a new learning objective for Multilingual neural machine translation (MNMT) based on distributionally robust optimization. We show how to practically optimize this objective for large translation corpora using an iterated best response scheme. Our method consistently outperforms strong baseline methods in terms of average and per-language performance under both many-to-one and one-to-many translation settings.
  arXiv  Detail & Related papers  (2021-09-09T03:48:35Z)
• Mixed-Lingual Pre-training for Cross-lingual Summarization [54.4823498438831]
  Cross-lingual Summarization aims at producing a summary in the target language for an article in the source language. We propose a solution based on mixed-lingual pre-training that leverages both cross-lingual tasks like translation and monolingual tasks like masked language models. Our model achieves an improvement of 2.82 (English to Chinese) and 1.15 (Chinese to English) ROUGE-1 scores over state-of-the-art results.
  arXiv  Detail & Related papers  (2020-10-18T00:21:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.