JsonGrinder is a collection of routines that facilitates conversion of JSON documents into structures used by
Mill.jl
project.
It provides schema estimation from data, extraction of various data types to numeric representation with reasonable defaults, suggestion of NN model structure based on data and interactive HTML visualization of estimated schema. For more details, see the documentation.
Watch our introductory talk from JuliaCon 2021
Installation
Run the following in REPL:
] add JsonGrinder
Getting Started
Four pointers to get you started:
- Examples: easy to understand JsonGrinder.jl and Mill.jl code across various domains
- The documentation
- The API Reference
- Dedicated examples repository containing few real-world usecases train on various datasets.
Citation
For citing, please use the following entry for the original paper:
@misc{mandlik2021milljl,
title={Mill.jl and JsonGrinder.jl: automated differentiable feature extraction for learning from raw JSON data},
author={Simon Mandlik and Matej Racinsky and Viliam Lisy and Tomas Pevny},
year={2021},
eprint={2105.09107},
archivePrefix={arXiv},
primaryClass={stat.ML}
}
and the following for this implementation (fill in the used version
):
@software{jsongrinder2019,
author = {Tomas Pevny and Matej Racinsky},
title = {JsonGrinder.jl: a flexible library for automated feature engineering and conversion of JSONs to Mill.jl structures},
url = {https://github.com/CTUAvastLab/JsonGrinder.jl},
version = {...},
}
Contribution guidelines
If you want to contribute to Mill.jl, be sure to review the contribution guidelines.
We use GitHub issues for tracking requests and bugs.