CSKG: The CommonSense Knowledge Graph

CSKG is a commonsense knowledge graph that combines seven popular sources into a consolidated representation:

• ATOMIC
• ConceptNet
• FrameNet
• Roget
• Visual Genome
• Wikidata (We use the Wikidata-CS subset)
• WordNet

CSKG is represented as a hyper-relational graph, by using the KGTK data model and file specification. Its creation is entirely supported by KGTK operations.

CSKG is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Getting started

Documentation

• https://cskg.readthedocs.io/en/latest/

Data

• CSKG
• CSKG with dimension info

Embeddings

• CSKG embeddings on google drive
• CSKG is integrated into PyKeen which allows easy computation of various (currently 25) state-of-the-art graph embeddings

More info

• Tutorial on Commonsense Knowledge Graphs (ISWC'20)
• Tutorial on Commonsense Knowledge Acquisition and Representation (AAAI'21)
• Workshop on Commonsense Knowledge Graphs (AAAI'21)

Consolidating your own CSKG

1. Setup your conda environment

conda create --name mowgli --file requirements.txt
conda activate mowgli

2. Download and store individual sources, except WordNet and FrameNet. By default, these should be stored in the input directory.

3. Download the mappings from this folder and place them inside the input directory

4. Customize and run create_cskg.sh.

How to cite

@article{ilievski2021cskg,
  title={CSKG: The CommonSense Knowledge Graph},
  author={Ilievski, Filip and Szekely, Pedro and Zhang, Bin},
  journal={Extended Semantic Web Conference (ESWC)},
  year={2021}
}