Skip to content

README.md

Latest commit

 

History

History
45 lines (28 loc) · 1.22 KB

README.md

File metadata and controls

45 lines (28 loc) · 1.22 KB
Raw

Dataset and Generation

Dataset is a directory saving 12 real-life and 8 synthetic datasets used in the experiments.

The following is the table explaining the details about the 20 datasets.

drawing

Prerequisites

  1. Download of the latest docker container and initiation.

Explanation About Datasets

Each dataset is comprised of three files

  • schema (.json)
  • positive instances (.jsonl)
  • negative instances (.jsonl)

The positive instances are

  • retrieved from public websites or APIs (for real-life datasets)
  • generated using the ground truth schemas (for synthetic datasets)

The schemas are

  • retrieved from public websites or manually translated into .json formats using the official documentations (for real-life datasets)
  • retrieved from the JSON Schema Store (syntehtic datasets)

The negative instances are

  • generated using the ground truth schemas

Negative Instances Generation

You can simply generate negative instances using the following codes.

This may take a few hours, so we recommend you to run this process with tmux.

cd SyntheticDatasetGen
./generateAllNegativeSamples.sh