Leveraging large language models for spelling correction in Turkish
The field of natural language processing (NLP) has rapidly progressed, particularly with the rise of large language models (LLMs), which enhance our understanding of the intrinsic structures of languages in a cross-linguistic manner for complex NLP tasks. However, commonly encountered misspellings i...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
PeerJ Inc.
2025-06-01
|
| Series: | PeerJ Computer Science |
| Subjects: | |
| Online Access: | https://peerj.com/articles/cs-2889.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849422562954575872 |
|---|---|
| author | Ceren Guzel Turhan |
| author_facet | Ceren Guzel Turhan |
| author_sort | Ceren Guzel Turhan |
| collection | DOAJ |
| description | The field of natural language processing (NLP) has rapidly progressed, particularly with the rise of large language models (LLMs), which enhance our understanding of the intrinsic structures of languages in a cross-linguistic manner for complex NLP tasks. However, commonly encountered misspellings in human-written texts adversely affect language understanding for LLMs for various NLP tasks as well as misspelling applications such as auto-proofreading and chatbots. Therefore, this study focuses on the task of spelling correction in the agglutinative language Turkish, where its nature makes spell correction significantly more challenging. To address this, the research introduces a novel dataset, referred to as NoisyWikiTr, to explore encoder-only models based on bidirectional encoder representations from transformers (BERT) and existing auto-correction tools. For the first time in this study, as far as is known, encoder-only models based on BERT are presented as subword prediction models, and encoder-decoder models based on text-cleaning (Text-to-Text Transfer Transformer) architecture are fine-tuned for this task in Turkish. A comprehensive comparison of these models highlights the advantages of context-based approaches over traditional, context-free auto-correction tools. The findings also reveal that among LLMs, a language-specific sequence-to-sequence model outperforms both cross-lingual sequence-to-sequence models and encoder-only models in handling realistic misspellings. |
| format | Article |
| id | doaj-art-3eb29958dd834526bd59a344f528f0a6 |
| institution | Kabale University |
| issn | 2376-5992 |
| language | English |
| publishDate | 2025-06-01 |
| publisher | PeerJ Inc. |
| record_format | Article |
| series | PeerJ Computer Science |
| spelling | doaj-art-3eb29958dd834526bd59a344f528f0a62025-08-20T03:31:02ZengPeerJ Inc.PeerJ Computer Science2376-59922025-06-0111e288910.7717/peerj-cs.2889Leveraging large language models for spelling correction in TurkishCeren Guzel Turhan0Department of Computer Engineering, Gazi University, Ankara, TurkeyThe field of natural language processing (NLP) has rapidly progressed, particularly with the rise of large language models (LLMs), which enhance our understanding of the intrinsic structures of languages in a cross-linguistic manner for complex NLP tasks. However, commonly encountered misspellings in human-written texts adversely affect language understanding for LLMs for various NLP tasks as well as misspelling applications such as auto-proofreading and chatbots. Therefore, this study focuses on the task of spelling correction in the agglutinative language Turkish, where its nature makes spell correction significantly more challenging. To address this, the research introduces a novel dataset, referred to as NoisyWikiTr, to explore encoder-only models based on bidirectional encoder representations from transformers (BERT) and existing auto-correction tools. For the first time in this study, as far as is known, encoder-only models based on BERT are presented as subword prediction models, and encoder-decoder models based on text-cleaning (Text-to-Text Transfer Transformer) architecture are fine-tuned for this task in Turkish. A comprehensive comparison of these models highlights the advantages of context-based approaches over traditional, context-free auto-correction tools. The findings also reveal that among LLMs, a language-specific sequence-to-sequence model outperforms both cross-lingual sequence-to-sequence models and encoder-only models in handling realistic misspellings.https://peerj.com/articles/cs-2889.pdfMasked language modelingEncoder-only LLMsEncoder-decoder sequence-to-sequence modelsSpell checkSpell correction |
| spellingShingle | Ceren Guzel Turhan Leveraging large language models for spelling correction in Turkish PeerJ Computer Science Masked language modeling Encoder-only LLMs Encoder-decoder sequence-to-sequence models Spell check Spell correction |
| title | Leveraging large language models for spelling correction in Turkish |
| title_full | Leveraging large language models for spelling correction in Turkish |
| title_fullStr | Leveraging large language models for spelling correction in Turkish |
| title_full_unstemmed | Leveraging large language models for spelling correction in Turkish |
| title_short | Leveraging large language models for spelling correction in Turkish |
| title_sort | leveraging large language models for spelling correction in turkish |
| topic | Masked language modeling Encoder-only LLMs Encoder-decoder sequence-to-sequence models Spell check Spell correction |
| url | https://peerj.com/articles/cs-2889.pdf |
| work_keys_str_mv | AT cerenguzelturhan leveraginglargelanguagemodelsforspellingcorrectioninturkish |