From Smør-re-brød to Subwords: Training LLMs on Danish, One Morpheme at a Time

• URL: http://arxiv.org/abs/2504.01540v1
• Date: Wed, 02 Apr 2025 09:26:02 GMT
• Title: From Smør-re-brød to Subwords: Training LLMs on Danish, One Morpheme at a Time
• Authors: Mikkel Wildner Kildeberg, Emil Allerslev Schledermann, Nicolaj Larsen, Rob van der Goot,
• Abstract summary: We leverage an annotated Danish morphological dataset to train a semi-supervised model for morphological segmentation.<n>We evaluate four distinct tokenizers, including two custom morphological tokenizers, by analyzing their performance intextly segmenting Danish words.<n>Our findings reveal that our custom-developed tokenizers substantially enhance morphological segmentation, achieving an F1 score of 58.84, compared to 39.28 achieved by a Danish BPE tokenizer.
• Score: 8.28573483085828
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: The best performing transformer-based language models use subword tokenization techniques, such as Byte-Pair-Encoding (BPE). However, these approaches often overlook linguistic principles, such as morphological segmentation, which we believe is fundamental for understanding language-specific word structure. In this study, we leverage an annotated Danish morphological dataset to train a semisupervised model for morphological segmentation, enabling the development of tokenizers optimized for Danish morphology. We evaluate four distinct tokenizers, including two custom morphological tokenizers, by analyzing their performance in morphologically segmenting Danish words. Additionally, we train two generative transformer models, \textit{CerebrasGPT-111M} and \textit{LLaMA-3.2 1B}, using these tokenizers and evaluate their downstream performance. Our findings reveal that our custom-developed tokenizers substantially enhance morphological segmentation, achieving an F1 score of 58.84, compared to 39.28 achieved by a Danish BPE tokenizer. In downstream tasks, models trained with our morphological tokenizers outperform those using BPE tokenizers across different evaluation metrics. These results highlight that incorporating Danish morphological segmentation strategies into tokenizers leads to improved performance in generative transformer models on Danish language

Related papers

• MAGNET: Improving the Multilingual Fairness of Language Models with Adaptive Gradient-Based Tokenization [81.83460411131931]
  In multilingual settings, non-Latin scripts and low-resource languages are usually disadvantaged in terms of language models' utility, efficiency, and cost. We propose multilingual adaptive gradient-based tokenization to reduce over-segmentation via adaptive gradient-based subword tokenization.
  arXiv  Detail & Related papers  (2024-07-11T18:59:21Z)
• Character-level NMT and language similarity [1.90365714903665]
  We explore the effectiveness of character-level neural machine translation for various levels of language similarity and size of the training dataset on translation between Czech and Croatian, German, Hungarian, Slovak, and Spanish. We evaluate the models using automatic MT metrics and show that translation between similar languages benefits from character-level input segmentation. We confirm previous findings that it is possible to close the gap by finetuning the already trained subword-level models to character-level.
  arXiv  Detail & Related papers  (2023-08-08T17:01:42Z)
• SelfSeg: A Self-supervised Sub-word Segmentation Method for Neural Machine Translation [51.881877192924414]
  Sub-word segmentation is an essential pre-processing step for Neural Machine Translation (NMT) This paper introduces SelfSeg, a self-supervised neural sub-word segmentation method. SelfSeg is much faster to train/decode and requires only monolingual dictionaries instead of parallel corpora.
  arXiv  Detail & Related papers  (2023-07-31T04:38:47Z)
• MorphPiece : A Linguistic Tokenizer for Large Language Models [3.8073142980733]
  I propose a linguistically motivated tokenization scheme, MorphPiece, which is based partly on morphological segmentation of the underlying text. A GPT-style causal language model trained on this tokenizer (called MorphGPT) shows comparable or superior performance on a variety of supervised and unsupervised NLP tasks.
  arXiv  Detail & Related papers  (2023-07-14T10:35:04Z)
• Effects of sub-word segmentation on performance of transformer language models [0.628122931748758]
  We compare GPT and BERT models trained with the statistical segmentation algorithm BPE vs. two unsupervised algorithms for morphological segmentation. We show that training with morphological segmentation allows the LMs to: 1. achieve lower perplexity, 2. converge more efficiently in terms of training time, and 3. achieve equivalent or better evaluation scores on downstream tasks.
  arXiv  Detail & Related papers  (2023-05-09T14:30:29Z)
• Structural Biases for Improving Transformers on Translation into Morphologically Rich Languages [120.74406230847904]
  TP-Transformer augments the traditional Transformer architecture to include an additional component to represent structure. The second method imbues structure at the data level by segmenting the data with morphological tokenization. We find that each of these two approaches allows the network to achieve better performance, but this improvement is dependent on the size of the dataset.
  arXiv  Detail & Related papers  (2022-08-11T22:42:24Z)
• Impact of Tokenization on Language Models: An Analysis for Turkish [2.4660652494309936]
  We train tokenizers and pretrain medium-sized language models using RoBERTa pretraining procedure on the Turkish split of the OSCAR corpus. Our experiments, supported by statistical tests, reveal that Morphological-level tokenizer has challenging performance with de facto tokenizers. We find that increasing the vocabulary size improves the performance of Morphological and Word-level tokenizers more than that of de facto tokenizers.
  arXiv  Detail & Related papers  (2022-04-19T12:01:46Z)
• BPE vs. Morphological Segmentation: A Case Study on Machine Translation of Four Polysynthetic Languages [38.5427201289742]
  We investigate a variety of supervised and unsupervised morphological segmentation methods for four polysynthetic languages. We compare the morphologically inspired segmentation methods against Byte-Pair s (BPEs) as inputs for machine translation. We show that for all language pairs except for Nahuatl, an unsupervised morphological segmentation algorithm outperforms BPEs consistently.
  arXiv  Detail & Related papers  (2022-03-16T21:27:20Z)
• A Massively Multilingual Analysis of Cross-linguality in Shared Embedding Space [61.18554842370824]
  In cross-lingual language models, representations for many different languages live in the same space. We compute a task-based measure of cross-lingual alignment in the form of bitext retrieval performance. We examine a range of linguistic, quasi-linguistic, and training-related features as potential predictors of these alignment metrics.
  arXiv  Detail & Related papers  (2021-09-13T21:05:37Z)
• Dynamic Programming Encoding for Subword Segmentation in Neural Machine Translation [80.38621085548013]
  This paper introduces Dynamic Programming (DPE) a new segmentation algorithm for tokenizing sentences into subword units. A mixed character-subword transformer is proposed, which enables exact log marginal likelihood estimation and exact MAP inference to find target segmentations.
  arXiv  Detail & Related papers  (2020-05-03T05:00:50Z)
• Byte Pair Encoding is Suboptimal for Language Model Pretraining [49.30780227162387]
  We analyze differences between unigram LM tokenization and byte-pair encoding (BPE) We find that the unigram LM tokenization method matches or outperforms BPE across downstream tasks and two languages. We hope that developers of future pretrained LMs will consider adopting the unigram LM method over the more prevalent BPE.
  arXiv  Detail & Related papers  (2020-04-07T21:21:06Z)

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.