Lightweight transformers for clinical natural language processing
Specialised pre-trained language models are becoming more frequent in Natural language Processing (NLP) since they can potentially outperform models trained on generic texts. BioBERT (Sanh et al., Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter. <i>arXiv preprint...
| Main Authors: | , , , , , , , |
|---|---|
| Other Authors: | |
| Format: | Journal article |
| Language: | English |
| Published: |
Cambridge University Press
2024
|
| _version_ | 1826316799319212032 |
|---|---|
| author | Rohanian, O Nouriborji, M Jauncey, H Kouchaki, S Nooralahzadeh, F Clifton, L Merson, L Clifton, DA |
| author2 | ISARIC Clinical Characterisation Group |
| author_facet | ISARIC Clinical Characterisation Group Rohanian, O Nouriborji, M Jauncey, H Kouchaki, S Nooralahzadeh, F Clifton, L Merson, L Clifton, DA |
| author_sort | Rohanian, O |
| collection | OXFORD |
| description | Specialised pre-trained language models are becoming more frequent in Natural language Processing (NLP) since they can potentially outperform models trained on generic texts. BioBERT (Sanh et al., Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter. <i>arXiv preprint arXiv: 1910.01108</i>, 2019) and BioClinicalBERT (Alsentzer et al., Publicly available clinical bert embeddings. In <i>Proceedings of the 2nd Clinical Natural Language Processing Workshop</i>, pp. 72–78, 2019) are two examples of such models that have shown promise in medical NLP tasks. Many of these models are overparametrised and resource-intensive, but thanks to techniques like knowledge distillation, it is possible to create smaller versions that perform almost as well as their larger counterparts. In this work, we specifically focus on development of compact language models for processing clinical texts (i.e. progress notes, discharge summaries, etc). We developed a number of efficient lightweight clinical transformers using knowledge distillation and continual learning, with the number of parameters ranging from million to million. These models performed comparably to larger models such as BioBERT and ClinicalBioBERT and significantly outperformed other compact models trained on general or biomedical data. Our extensive evaluation was done across several standard datasets and covered a wide range of clinical text-mining tasks, including natural language inference, relation extraction, named entity recognition and sequence classification. To our knowledge, this is the first comprehensive study specifically focused on creating efficient and compact transformers for clinical NLP tasks. The models and code used in this study can be found on our Huggingface profile at https://huggingface.co/nlpie and Github page at https://github.com/nlpie-research/Lightweight-Clinical-Transformers, respectively, promoting reproducibility of our results. |
| first_indexed | 2024-04-09T03:56:07Z |
| format | Journal article |
| id | oxford-uuid:86fcc94c-4479-49be-84c4-04e16ce0591f |
| institution | University of Oxford |
| language | English |
| last_indexed | 2025-02-19T04:28:32Z |
| publishDate | 2024 |
| publisher | Cambridge University Press |
| record_format | dspace |
| spelling | oxford-uuid:86fcc94c-4479-49be-84c4-04e16ce0591f2024-12-12T17:12:22ZLightweight transformers for clinical natural language processingJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:86fcc94c-4479-49be-84c4-04e16ce0591fEnglishSymplectic ElementsCambridge University Press2024Rohanian, ONouriborji, MJauncey, HKouchaki, SNooralahzadeh, FClifton, LMerson, LClifton, DAISARIC Clinical Characterisation GroupSpecialised pre-trained language models are becoming more frequent in Natural language Processing (NLP) since they can potentially outperform models trained on generic texts. BioBERT (Sanh et al., Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter. <i>arXiv preprint arXiv: 1910.01108</i>, 2019) and BioClinicalBERT (Alsentzer et al., Publicly available clinical bert embeddings. In <i>Proceedings of the 2nd Clinical Natural Language Processing Workshop</i>, pp. 72–78, 2019) are two examples of such models that have shown promise in medical NLP tasks. Many of these models are overparametrised and resource-intensive, but thanks to techniques like knowledge distillation, it is possible to create smaller versions that perform almost as well as their larger counterparts. In this work, we specifically focus on development of compact language models for processing clinical texts (i.e. progress notes, discharge summaries, etc). We developed a number of efficient lightweight clinical transformers using knowledge distillation and continual learning, with the number of parameters ranging from million to million. These models performed comparably to larger models such as BioBERT and ClinicalBioBERT and significantly outperformed other compact models trained on general or biomedical data. Our extensive evaluation was done across several standard datasets and covered a wide range of clinical text-mining tasks, including natural language inference, relation extraction, named entity recognition and sequence classification. To our knowledge, this is the first comprehensive study specifically focused on creating efficient and compact transformers for clinical NLP tasks. The models and code used in this study can be found on our Huggingface profile at https://huggingface.co/nlpie and Github page at https://github.com/nlpie-research/Lightweight-Clinical-Transformers, respectively, promoting reproducibility of our results. |
| spellingShingle | Rohanian, O Nouriborji, M Jauncey, H Kouchaki, S Nooralahzadeh, F Clifton, L Merson, L Clifton, DA Lightweight transformers for clinical natural language processing |
| title | Lightweight transformers for clinical natural language processing |
| title_full | Lightweight transformers for clinical natural language processing |
| title_fullStr | Lightweight transformers for clinical natural language processing |
| title_full_unstemmed | Lightweight transformers for clinical natural language processing |
| title_short | Lightweight transformers for clinical natural language processing |
| title_sort | lightweight transformers for clinical natural language processing |
| work_keys_str_mv | AT rohaniano lightweighttransformersforclinicalnaturallanguageprocessing AT nouriborjim lightweighttransformersforclinicalnaturallanguageprocessing AT jaunceyh lightweighttransformersforclinicalnaturallanguageprocessing AT kouchakis lightweighttransformersforclinicalnaturallanguageprocessing AT nooralahzadehf lightweighttransformersforclinicalnaturallanguageprocessing AT cliftonl lightweighttransformersforclinicalnaturallanguageprocessing AT mersonl lightweighttransformersforclinicalnaturallanguageprocessing AT cliftonda lightweighttransformersforclinicalnaturallanguageprocessing |