AUTOMATIC GENERATION OF ELECTRONIC CLINICAL RECORDS THROUGH NLP-BASED ANALYSIS OF PATIENT–PHYSICIAN CONVERSATIONS
Main Article Content
Abstract
Transforming a patient–physician conversation into an electronic clinical record requires the integration of automatic speech recognition, speaker diarization, clinical information extraction, and factually constrained generation. This study synthesizes research published between 2019 and 2026 on clinical NLP, medical ASR, dialogue-to-note generation, and ambient scribing, and proposes a conceptual architecture adapted to Uzbek. Published benchmark values were extracted together with their datasets and evaluation settings. On the English MedDictate benchmark, real-time Whisper achieved a word error rate of 40.3% and medical-term recall of 90.0%, whereas the domain-specialized Symphony system reported 4.1% WER and 93.9% medical-term recall. On an Uzbek named-entity recognition benchmark, BERTbek News-Big achieved an F1 score of 78.69%, compared with 75.14% for multilingual BERT. A 2026 MTS-Dialog study reported ROUGE-1 of 0.4574 and BERTScore-F1 of 0.9060 for Llama-3-8B, while Meditron-7B obtained the highest ROUGE-L of 0.3667. In MEDIQA-Chat Task B, the GPT-4-based WangLab approach reached ROUGE-1 of 0.6141 for full-note generation. A clinical implementation study found that ambient scribing reduced time in notes from 10.3 to 8.2 minutes per appointment and after-hours documentation from 50.6 to 35.4 minutes per workday. The synthesis supports an Uzbek clinical documentation system combining domain-adapted ASR, medical entity and relation extraction, evidence-linked structured generation, and mandatory physician review. Values reported in this article are published results from distinct studies and must not be interpreted as a single head-to-head experiment.
Article Details
References
Moy A.J., Schwartz J.M., Chen R., et al. Measurement of clinical documentation burden among physicians and nurses using electronic health records: a scoping review. Journal of the American Medical Informatics Association. 2021;28(5):998–1008. doi:10.1093/jamia/ocaa325.
Blackley S.V., Huynh J., Wang L., Korach Z., Zhou L. Speech recognition for clinical documentation from 1990 to 2018: a systematic review. Journal of the American Medical Informatics Association. 2019;26(4):324–338. doi:10.1093/jamia/ocy179.
Radford A., Kim J.W., Xu T., Brockman G., McLeavey C., Sutskever I. Robust Speech Recognition via Large-Scale Weak Supervision. Proceedings of ICML. PMLR 202; 2023:28492–28518.
Baevski A., Zhou H., Mohamed A., Auli M. wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations. Advances in Neural Information Processing Systems. 2020;33:12449–12460.
Babu A., Wang C., Tjandra A., et al. XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale. Interspeech 2022:2278–2282. doi:10.21437/Interspeech.2022-143.