Linked Papers With Code

Linked Papers With Code (LPWC) is an RDF knowledge graph that comprehensively models the research field of machine learning. It contains information about almost 400,000 machine learning publications, including the tasks addressed, the datasets utilized, the methods implemented, and the evaluations conducted, along with their results. The data set is based on Papers With Code. The underlying dataset snapshot of Papers With Code are regenerated daily. With the scripts provided in this repository, the LPWC knowledeg graph can be re-generated based on the snapshot. Furthermore, we provide knowledge graph embeddings for entities and relations represented in LPWC.

LPWC is available at https://linkedpaperswithcode.com.

For an overview of LPWC see our ISWC2023 Poster.

Schema of Linked Papers With Code

Knowledge Graph Construction

Data transformation

To construct the Linked Papers With Code Knowledge Graph based on the Paperswithcode dataset snapshot we use the following python scripts:

1. 01_papers.py
2. 02_conferences.py
3. 03_methods.py
4. 04_papers-code-and-datasets.py
5. 05_evaluation-tables.py
6. 06_validation-and-formats.py

Entity Linking

The scripts for Linking the Entities of Linked Papers With Code to SemOpenAlex and Wikidata are in the folder entity-linking-scripts.

Knowledge Graph Embeddings

The scripts for generating the Knowledge Graph Embeddings and the Evaluation Results of the different Knowledge Graph Embeddings Models are in the folder embeddings-generation.

Reference & More Information

More Information about LPWC and its creation can be found in our paper.