File size: 1,046 Bytes
21562da
9b65513
21562da
9b65513
 
 
 
 
 
 
 
 
 
 
 
 
21562da
9b65513
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
---
language: id
license: gpl-3.0
task_categories:
- automatic-speech-recognition
tags:
- medical
dataset_info:
  splits:
  - name: train
    num_bytes: 2705299088
    num_examples: 360
  - name: test
    num_bytes: 479374176
    num_examples: 360
  download_size: 3184673264
---
# Building MEDISCO: Indonesian Speech Corpus for Medical Domain
The dataset was published in the following paper:
> Building MEDISCO: Indonesian Speech Corpus for Medical Domain ([PDF](https://mrqorib.github.io/assets/pdf/MEDISCO.pdf) | [IEEEXplore](https://ieeexplore.ieee.org/abstract/document/8629259)) <br>
> [Muhammad Reza Qorib](https://mrqorib.github.io/) and [Mirna Adriani](https://cs.ui.ac.id/en/personnel/mirna-adriani/) <br>
> 2018 International Conference on Asian Language Processing (IALP)

Please look for the raw files (inside the "Files and versions" tab) as the dataset viewer parsed by Huggingface does not show the text transcript.
Please direct any questions to mrqorib[at]u.nus.edu as the email addresses in the paper PDF are no longer active.