Can Large Language Models Imitate Human Speech for Clinical Assessment? LLM-Driven Data Augmentation for Cognitive Score Prediction
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914569526444032 |
|---|---|
| author | Ketir, Si-Belkacem Yamine Tamayo, Lenard Paulo Hisada, Shohei Peng, Shaowen Wakamiya, Shoko Aramaki, Eiji |
| author_facet | Ketir, Si-Belkacem Yamine Tamayo, Lenard Paulo Hisada, Shohei Peng, Shaowen Wakamiya, Shoko Aramaki, Eiji |
| contents | Accurate assessment of cognitive decline from spontaneous speech remains challenging due to limited dataset size and class imbalance. In this work, we propose a large language model (LLM)-driven data augmentation framework to improve the prediction of cognitive scores from speech. Experiments are conducted on a Japanese corpus in which each participant provides both a spontaneous oral narrative and a written response to the same clinical prompt. The written responses serve as semantic anchors to generate multiple oral-like monologues in different styles using GPT-5. We then predict Hasegawa Dementia Scale scores, a widely used cognitive screening tool in Japan, using a Partial Least Squares regression model trained on Sentence-BERT speech embeddings. We investigate two augmentation strategies: random class-balanced selection, which yields moderate but unstable improvements, and similarity-guided class-balanced selection. The latter prioritizes semantically close synthetic samples, leading to more consistent improvements and substantially reducing prediction error for minority low-score participants while maintaining performance for the majority group. Overall, our findings demonstrate the potential of semantically guided LLM-driven augmentation as a principled approach for addressing class imbalance and improving data efficiency in clinical speech analysis. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_16077 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Can Large Language Models Imitate Human Speech for Clinical Assessment? LLM-Driven Data Augmentation for Cognitive Score Prediction Ketir, Si-Belkacem Yamine Tamayo, Lenard Paulo Hisada, Shohei Peng, Shaowen Wakamiya, Shoko Aramaki, Eiji Computation and Language I.2.7; J.3 Accurate assessment of cognitive decline from spontaneous speech remains challenging due to limited dataset size and class imbalance. In this work, we propose a large language model (LLM)-driven data augmentation framework to improve the prediction of cognitive scores from speech. Experiments are conducted on a Japanese corpus in which each participant provides both a spontaneous oral narrative and a written response to the same clinical prompt. The written responses serve as semantic anchors to generate multiple oral-like monologues in different styles using GPT-5. We then predict Hasegawa Dementia Scale scores, a widely used cognitive screening tool in Japan, using a Partial Least Squares regression model trained on Sentence-BERT speech embeddings. We investigate two augmentation strategies: random class-balanced selection, which yields moderate but unstable improvements, and similarity-guided class-balanced selection. The latter prioritizes semantically close synthetic samples, leading to more consistent improvements and substantially reducing prediction error for minority low-score participants while maintaining performance for the majority group. Overall, our findings demonstrate the potential of semantically guided LLM-driven augmentation as a principled approach for addressing class imbalance and improving data efficiency in clinical speech analysis. |
| title | Can Large Language Models Imitate Human Speech for Clinical Assessment? LLM-Driven Data Augmentation for Cognitive Score Prediction |
| topic | Computation and Language I.2.7; J.3 |
| url | https://arxiv.org/abs/2605.16077 |