Shih-Yen Hou

Shih-Yen Hou

Auto-generated profile, based on publicly available information

Shih-Yen Hou

Artificial Intelligence and Robotics Innovation Center, China Medical University Hospital, China Medical University, No. 2, Yude Rd, North Dist, Taichung, 404327, Taiwan, 886 0422052121 ext 12584

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 August 2023 in JMIR Medical Education
Verified
Journal Article
Open access
HHHsing-Yu Hsu
Hsing-Yu Hsu
KHKai-Cheng Hsu
Kai-Cheng Hsu
SHShih-Yen Hou
Shih-Yen Hou
CWChing-Lung Wu
Ching-Lung Wu
YHYow-Wen Hsieh
Yow-Wen Hsieh
YCYih-Dih Cheng
Yih-Dih Cheng
Background: Since OpenAI released ChatGPT, with its strong capability in handling natural tasks and its user-friendly interface, it has garnered significant attention. Objective: A prospective analysis is required to evaluate the accuracy and appropriateness of medication consultation responses generated by ChatGPT. Methods: A prospective cross-sectional study was conducted by the pharmacy department of a medical center in Taiwan. The test data set comprised retrospective medication consultation questions collected from February 1, 2023, to February 28, 2023, along with common questions about drug-herb interactions. Two distinct sets of questions were tested: real-world medication consultation questions and common questions about interactions between traditional Chinese and Western medicines. We used the conventional double-review mechanism. The appropriateness of each response from ChatGPT was assessed by 2 experienced pharmacists. In the event of a discrepancy between the assessments, a third pharmacist stepped in to make the final decision. Results: Of 293 real-world medication consultation questions, a random selection of 80 was used to evaluate ChatGPT’s performance. ChatGPT exhibited a higher appropriateness rate in responding to public medication consultation questions compared to those asked by health care providers in a hospital setting (31/51, 61% vs 20/51, 39%; P=.01). Conclusions: The findings from this study suggest that ChatGPT could potentially be used for answering basic medication consultation questions. Our analysis of the erroneous information allowed us to identify potential medical risks associated with certain questions; this problem deserves our close attention.

7 December 2022 in JMIR Nursing
Verified
Journal Article
Open access
SHShih-Yen Hou
Shih-Yen Hou
YWYa-Lun Wu
Ya-Lun Wu
KCKai-Ching Chen
Kai-Ching Chen
CTChang Ting-An
Chang Ting-An
YHYi-Min Hsu
Yi-Min Hsu
KHKai-Cheng Hsu
Kai-Cheng Hsu
Background: Taiwan has insufficient nursing resources due to the high turnover rate of health care providers. Therefore, reducing the heavy workload of these employees is essential. Herein, speech transcription, which has various potential clinical applications, was employed for the documentation of nursing records. The requirement of including only one speaker per transcription facilitated data collection and system development. Moreover, authorization from patients was unnecessary. Objective: The aim of this study was to construct a speech recognition system for nursing records such that health care providers can complete nursing records without typing or with only a few edits. Methods: Nursing records in Taiwan are mainly written in Mandarin, with technical terms and abbreviations presented in both Mandarin and English. Therefore, the training set consisted of English code-switching information. Next, transfer learning (TL) and meta-TL (MTL) methods, which perform favorably in code-switching scenarios, were applied. Results: As of September 2021, the China Medical University Hospital Artificial Intelligence Speech (CMaiSpeech) data set was established by manually annotating approximately 100 hours of recordings from 525 speakers. The word error rate (WER) of the benchmark model of syllable-based TL was 29.54% in code-switching. The WER of the proposed model of syllable-based MTL was 22.20% in code-switching. The test set comprised 17,247 words. Moreover, in a clinical case, the proposed model of syllable-based MTL yielded a WER of 31.06% in code-switching. The clinical test set contained 1159 words. Conclusions: This paper has two main contributions. First, the CMaiSpeech data set—a Mandarin-English corpus—has been established. Health care providers in Taiwan are often compelled to use a mixture of Mandarin and English in nursing records. Second, an automatic speech recognition system for nursing record document conversion was proposed. The proposed system can shorten the work handover time and further reduce the workload of health care providers.

5 June 2026 in JMIR Nursing
Verified
Journal Article
Open access
MSMeng-Han Su
Meng-Han Su
WWWei-Chun Wang
Wei-Chun Wang
YHYi-Min Hsu
Yi-Min Hsu
SHShih-Yen Hou
Shih-Yen Hou
SCSu-Jung Chuang
Su-Jung Chuang
SCShih-Sheng Chang
Shih-Sheng Chang
Background: For clinical nurses, manually entering information into hospital information systems (HISs) remains time-consuming and prone to omissions. Although speech recognition can reduce the need for manual entry, its use in clinical settings has historically been limited by code-switching, medical terminology, and noisy ward environments. Recent advances in customized automatic speech recognition (ASR) and large language models (LLMs) now make speech-based, structured documentation aligned with nursing frameworks such as DART (data, action, response, and teaching) increasingly feasible. Objective: This study developed and evaluated an integrated ASR and LLM system that transforms spoken nursing input into structured DART notes and evaluated its accuracy, usability, and clinical feasibility within HIS workflows. Methods: A code-switching nursing speech corpus from emergency and ward settings was used to fine-tune the Whisper large-v2 model with parameter-efficient adaptation. The LLM generated schema-constrained DART records from ASR transcripts, which were verified by nurses before being uploaded to the corresponding HIS fields. Evaluation included mixed error rate for ASR accuracy, F 1 -scores, and agreement statistics for DART classification, hallucination assessments based on factual correctness, and analysis of nurse feedback on system use. Results: The fine-tuned ASR model reduced the mixed error rate from 44.79% to 6.67%. DART generation achieved a macroaveraged F 1 -score of 0.82 (95% CI 0.80‐0.84) and met the noninferiority margin relative to human transcripts (δ=−0.04). The hallucination rate was 2.51%. During deployment, the monthly volume of valid nursing notes generated through voluntary use of the ASR system increased from 32,724 to 65,417, where each note represented a single documentation entry generated per patient care episode. Among 120 participating nurses, 91 (75.8%) reported reduced workload and improved completeness. Conclusions: The integrated ASR and LLM system was feasible and showed strong performance, with good acceptance among clinical nurses. It reduced the manual documentation burden, improved record completeness, and demonstrated the value of an ASR- and LLM-supported workflow for nursing documentation.

24 April 2023 in JMIR Preprints
Verified
Preprint
Open access
HHHsing-Yu Hsu
Hsing-Yu Hsu
KHKai-Cheng Hsu
Kai-Cheng Hsu
SHShih-Yen Hou
Shih-Yen Hou
CWChing-Lung Wu
Ching-Lung Wu
YHYow-Wen Hsieh
Yow-Wen Hsieh
YCYih-Dih Cheng
Yih-Dih Cheng
BACKGROUND Since OpenAI released ChatGPT, with its strong capability in handling natural tasks and its user-friendly interface, it has garnered significant attention. OBJECTIVE A prospective analysis is required to evaluate the accuracy and appropriateness of medication consultation responses generated by ChatGPT. METHODS A prospective cross-sectional study was conducted by the pharmacy department of a medical center in Taiwan. The test data set comprised retrospective medication consultation questions collected from February 1, 2023, to February 28, 2023, along with common questions about drug-herb interactions. Two distinct sets of questions were tested: real-world medication consultation questions and common questions about interactions between traditional Chinese and Western medicines. We used the conventional double-review mechanism. The appropriateness of each response from ChatGPT was assessed by 2 experienced pharmacists. In the event of a discrepancy between the assessments, a third pharmacist stepped in to make the final decision. RESULTS Of 293 real-world medication consultation questions, a random selection of 80 was used to evaluate ChatGPT’s performance. ChatGPT exhibited a higher appropriateness rate in responding to public medication consultation questions compared to those asked by health care providers in a hospital setting (31/51, 61% vs 20/51, 39%; P=.01). CONCLUSIONS The findings from this study suggest that ChatGPT could potentially be used for answering basic medication consultation questions. Our analysis of the erroneous information allowed us to identify potential medical risks associated with certain questions; this problem deserves our close attention.

27 November 2025 in JMIR Preprints
Verified
Preprint
Open access
MSMeng-Han Su
Meng-Han Su
WWWei-Chun Wang
Wei-Chun Wang
YHYi-Min Hsu
Yi-Min Hsu
SHShih-Yen Hou
Shih-Yen Hou
SCSu-Jung Chuang
Su-Jung Chuang
SCShih-Sheng Chang
Shih-Sheng Chang
BACKGROUND For clinical nurses, manually entering information into hospital information systems (HIS) remains time consuming and prone to omissions. Although speech recognition can reduce the need for manual entry, its use in clinical settings has historically been limited by code switching, medical terminology, and noisy ward environments. Recent advances in customized automatic speech recognition (ASR) and large language models (LLMs) now make speech based, structured documentation aligned with nursing frameworks such as D.A.R.T. (DaThis study developed and evaluated an integrated ASR and LLM system that transforms spoken nursing input into structured D.A.R.T. notes, evaluating its accuracy, usability, and clinical feasibility within HIS workflows.ta, Action, Response, Teaching) increasingly feasible. OBJECTIVE This study developed and evaluated an integrated ASR and LLM system that transforms spoken nursing input into structured D.A.R.T. notes, evaluating its accuracy, usability, and clinical feasibility within HIS workflows. METHODS A code-switching nursing speech corpus from emergency and ward settings was used to fine tune the Whisper Large v2 model with parameter efficient adaptation[1]. The LLM generated schema constrained D.A.R.T. records from ASR transcripts, which were verified by nurses before being uploaded to the corresponding HIS fields. Evaluation included mixed error rate (MER) for ASR accuracy, F1 scores and agreement statistics for D.A.R.T. classification, hallucination assessments based on factual correctness, and analysis of nurse feedback on system use. RESULTS The fine tuned ASR model reduced the mixed error rate from 31.65% to 6.6%. D.A.R.T. generation achieved an F1 score of 0.82 and met the non-inferiority margin relative to human transcripts (Δ = −0.04).The hallucination rate was 2.51%. During deployment, monthly documentation volumes increased from 32,724 to 65,417, and 76% of the 120 participating nurses reported reduced workload and improved completeness. CONCLUSIONS The integrated ASR and LLM system was feasible and showed strong performance, with good acceptance among clinical nurses. It reduced manual documentation burden, improved record completeness, and demonstrated the value of an ASR and LLM supported workflow for nursing documentation.