Skip to content

These are different files I created to do different tasks when I was working on creating ASR model for mTEDx dataset.

License

Notifications You must be signed in to change notification settings

Anwarvic/mTEDx_auxiliary

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

mTEDx_auxiliary

These are different files I created to do different tasks when I was working on creating ASR model for mTEDx dataset. The following is a description of the different files you can find in this repo:

  • download_mted.sh: A bash script to download the ASR part of the mTEDx dataset.
  • process_mtedx: A file to process the mTEDx dataset. By processing, I mean:
    • Split the audio of the ted talks found in the datasets into shorter segments that can be used for training ASR models.
    • Normalizes the audio files to a sample rate of16000 and number of channels of 1.
    • Align between the audio segments and the text.
  • stats.xlsx: An excel file gathering all the details of the data found in the dataset.

How to download the data

You can download mTEDx data by running the following bash script:

bash download_mtedx.sh

How the data should look like

After downloading the audio data, you need to process the audio data and split the whole talks into different segments and align it with the correct text. All of that can be done by running this python script like so:

python process_mtedx.py \
  --in [IN_DATA_PATH] \
  --out [OUT_DATA_PATH] \
  --langs [LANGS] \
  --groups [GROUPS]

For example:

python process_mtedx.py \
  --in /scratch/1/user/manwar/data/mTEDx \
  --out /scratch/1/user/manwar/data/mTEDx_wav \
  --langs ar el es de fr it pt ru\
  --groups test valid train

This will process all audio data and organize it in the following tree:

mTEDx
├── ar
│   ├── test
│   ├── test.json
│   ├── train
│   ├── train.json
│   ├── valid
│   └── valid.json
├── de
│   ├── test
│   ├── test.json
│   ├── train
│   ├── train.json
│   ├── valid
│   └── valid.json
├── el
│   ├── test
│   ├── test.json
│   ├── train
│   ├── train.json
│   ├── valid
│   └── valid.json
├── es
│   ├── test
│   ├── test.json
│   ├── train
│   ├── train.json
│   ├── valid
│   └── valid.json
├── fr
│   ├── test
│   ├── test.json
│   ├── train
│   ├── train.json
│   ├── valid
│   └── valid.json
├── it
│   ├── test
│   ├── test.json
│   ├── train
│   ├── train.json
│   ├── valid
│   └── valid.json
├── pt
│   ├── test
│   ├── test.json
│   ├── train
│   ├── train.json
│   ├── valid
│   └── valid.json
└── ru
    ├── test
    ├── test.json
    ├── train
    ├── train.json
    ├── valid
    └── valid.json

About

These are different files I created to do different tasks when I was working on creating ASR model for mTEDx dataset.

Topics

Resources

License

Stars

Watchers

Forks