configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
dataset_info:
features:
- name: audio
struct:
- name: array
sequence:
sequence: float32
- name: path
dtype: string
- name: sampling_rate
dtype: int64
- name: sentence
dtype: string
splits:
- name: train
num_bytes: 3128740048
num_examples: 5328
- name: test
num_bytes: 776455056
num_examples: 1333
download_size: 3882364624
dataset_size: 3905195104
license: apache-2.0
task_categories:
- automatic-speech-recognition
language:
- en
tags:
- medical
size_categories:
- 1K<n<10K
Data Source
Kaggle Medical Speech, Transcription, and Intent
Context
8.5 hours of audio utterances paired with text for common medical symptoms.
Content
This data contains thousands of audio utterances for common medical symptoms like “knee pain” or “headache,” totaling more than 8 hours in aggregate. Each utterance was created by individual human contributors based on a given symptom. These audio snippets can be used to train conversational agents in the medical field.
This Figure Eight dataset was created via a multi-job workflow. The first involved contributors writing text phrases to describe symptoms given. For example, for “headache,” a contributor might write “I need help with my migraines.” Subsequent jobs captured audio utterances for accepted text strings.
Note that some of the labels are incorrect and some of the audio files have poor quality. I would recommend cleaning the dataset before training any machine learning models.
This dataset contains both the audio utterances and corresponding transcriptions.
What's new
*The data is clean from all columns except for the file_path and phrase
*All Audios are loaded into the DatasetDict as an 1D array, float32
*All Audios are resampled into 16K
*The new structure :
train = {
'audio': {
'path': file_path, the mp3 files is not included here, please visit the kaggle to dowload em
'array': waveform_np,
'sampling_rate': 16000
},
'sentence': the text transcription
}