Multilingual multitask medical speech processing, ASR, NER, and summarization.
Multilingual medical speech recognition, NER, and real-time conversation summarization datasets and models.
MultiMed provides comprehensive open-source datasets, pre-trained models, and benchmarks for medical speech processing across multiple languages. It is designed for researchers and engineers building automated speech recognition, named entity recognition, and summarization tools for clinical environments.
It bundles research contributions including VietMed for speech recognition, VietMed-NER for spoken named entity recognition across 18 entity types, VietMed-Sum for real-time medical conversation summarization, and MultiMed for multilingual medical ASR in five languages. The framework supplies Attention Encoder-Decoder models, pre-trained w2v2-Viet and XLSR-53-Viet weights, and codebases spanning major NLP and speech conferences (LREC-COLING, Interspeech, NAACL, ACL).
It suits researchers, machine learning engineers, and healthcare technologists developing clinical voice interfaces or medical NLP pipelines. It does not suit clinicians looking for a ready-to-use end-user medical transcription app without technical deployment capability.
Includes 16 hours of labeled and 1000 hours of unlabeled medical speech covering all ICD-10 disease groups.
Features VietMed-NER for extracting 18 distinct medical entity types from speech audio and transcripts.
Generates local summaries after every N utterances and global summaries at conversation end.
Provides end-to-end medical ASR models spanning Vietnamese, English, German, French, and Mandarin Chinese.
The code, data, and models are provided freely via the repository links, and the only cost is running the required hardware infrastructure yourself.
873 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
These reviewed AI tools do the same job instantly — from our scored directory.
The core MultiMed ASR models support five languages: Vietnamese, English, German, French, and Mandarin Chinese.
Yes, by translating the transcripts, the VietMed-NER dataset can be utilized for text-based named entity recognition in other languages.
Yes, the repository provides links to pre-trained models such as w2v2-Viet, XLSR-53-Viet, and fine-tuned medical ASR models.
See Zekai first in Google