Please add Parrot radiology asr en

[ysdede]

This dataset contains synthetic English radiology report speech recordings paired with their transcriptions, created for training Automatic Speech Recognition (ASR) models on medical radiology domain text.

The source text is derived from the PARROT v1.0 dataset, a multilingual collection of fictional radiology reports written by expert radiologists from 21 countries.

Language: English (source reports from 14 languages, translated to English)
Domain: Medical/Radiology
Task: Automatic Speech Recognition (ASR)
Total Audio: Approximately 55 hours
Total Samples: 9,484 audio segments
Audio Format: MP3 (VBR Quality 5, approximately 64kbps, 16kHz, mono)
Generation Method: Kokoro TTS (82M parameter model, v0.1.0)
Source Dataset: PARROT v1.0 (2,658 fictional radiology reports)
Dataset Format: Parquet

https://huggingface.co/datasets/ysdede/parrot-radiology-asr-en