Improve datasets loaders documentation

Accenture / AmpliGraph

Python library for Representation Learning on Knowledge Graphs https://docs.ampligraph.org

Apache License 2.0

2.15k stars 250 forks source link

Closed lukostaz closed 5 years ago

lukostaz commented 5 years ago

Description We must add in docstrings:

stats on how many tripels in each split, how many distinct entities and relations
warning boxes for datasets with missing entities (FB15k-237, WN18RR)
references to where the dataset was first proposed, and from where we downloaded it.

A summary table in the "datasets" section would also help. Something like this:

chanlevan commented 5 years ago

In performance page, I have added information of triples with unseen entities are removed in wn18rr and fb15k-237