Name		Name	Last commit message	Last commit date
parent directory ..
01_the_ACME_dataset.ipynb		01_the_ACME_dataset.ipynb
02_dataset_creation.ipynb		02_dataset_creation.ipynb
03_degradation_functions.ipynb		03_degradation_functions.ipynb
04_data_parsers_and_degrader.ipynb		04_data_parsers_and_degrader.ipynb
05_AMT_error_matching.ipynb		05_AMT_error_matching.ipynb
06_training_and_evalutation.ipynb		06_training_and_evalutation.ipynb
README.md		README.md
utils.py		utils.py

README.md

MDTK Documentation

This directory contains documentation with examples for the toolkit. This includes:

The ACME dataset
- an introduction and
- a description of the dataset used in our paper
Dataset creation
- how to create your own ACME datasets
Degradation functions
- an introduction to the available functions, and
- their parameters
Data parsers and the degrader class
- How to parse data and provide it to, for example, pytorch models
- How to augment a dataset with degradations on-the-fly
Matching errors with your AMT system
- How to generate data which matches the output of your AMT system
Reproducing results from the paper
- script to perform training & evaluation to reproduce paper results provided

If you are interested in augmenting the data to train a model which cleans the output of your AMT system:

Read how to match the errors of your AMT system, then either
1. how to create a dataset with parameters to match your AMT system's errors, or
2. degrade data on the fly
Train a model on the degraded data to fix the errors