mediTalk-mm-rdy
收藏资源简介:
mediTalk-mm-rdy是一个专为缅甸语临床对话自动语音识别构建的高质量医疗语音语料库,由泰国国家电子与计算机技术中心与缅甸语言理解实验室等机构联合创建。该语料库包含21,724条语音样本,总时长约28小时,均由9名缅甸语母语者基于PLAB 2临床评估手册的翻译文本录制,并经过严格的翻译验证与音频质量筛选。创建过程中,还通过模拟房间声学环境生成多样化训练数据,以增强模型鲁棒性。该数据集旨在填补缅甸语医疗语音资源的空白,为低资源场景下的ASR模型微调与鲁棒性评估提供基准支持。
MediTalk-mm-rdy is a high-quality medical speech corpus specifically constructed for Burmese clinical dialogue automatic speech recognition (ASR), jointly created by institutions including the National Electronics and Computer Technology Center of Thailand and the Myanmar Language Understanding Laboratory. This corpus contains 21,724 speech samples with a total duration of approximately 28 hours. All recordings were made by 9 native Burmese speakers based on translated texts from the PLAB 2 Clinical Assessment Handbook, and underwent strict translation validation and audio quality screening. During the corpus development process, diverse training data were also generated by simulating room acoustic environments to enhance model robustness. This dataset aims to fill the gap in Burmese medical speech resources, providing benchmark support for ASR model fine-tuning and robustness evaluation in low-resource scenarios.

- 1myMediWhisper: Construction of Burmese Medical Speech Corpus and Whisper Fine-Tuning for Clinical Dialogue ASR泰国国家电子与计算机技术中心; 缅甸语言理解实验室; 泰国国王科技大学; 泰国国王理工学院 · 2026年



