Wikidata Nearest Neighbor, Nearest Relation

Utilizes facebook research's BigGraph to find closest neighbors;

search.py includes a closest neighbor function, which takes in a wikidata entity and outputs the wikidata entities that are most closely related to the input.

It requires running the id_to_index and tsv_to_np code first as those will give you the files that search.py uses to run a fast and efficient algorithm.

How to use it:

1: Download wikidata_translation_v1_names.json:

https://dl.fbaipublicfiles.com/torchbiggraph/wikidata_translation_v1_names.json.gz

This file allows the search.ipynb code to map wikidata entities to names.

2: Run the id_to_index.ipynb code:

This converts the data from wikidata_translation_v1_names.json into .pkl files separated such that search.ipynb can process them.

3: Open search.ipynb:

Download the libraries that are imported, most notably, the installation of faiss can be found here: https://github.com/facebookresearch/faiss

Choose to either run the closest neighbor code, or the closest relations code. For each function, the required specified parameters are commented within.

Note: As of now the closest relations function is not working, but this is because the graph embeddings from fb research do not support it yet.

Relevant threads:

facebookresearch/PyTorch-BigGraph#13

facebookresearch/PyTorch-BigGraph#47

Note: tsv_to_np.py can be used to convert tsv files to np arrays. This is useful if you use another embedding besides wikidata, such as the FB15k data which is also available from Facebook Reasearch's BigGraph

Citations:

https://github.com/facebookresearch/PyTorch-BigGraph

@inproceedings{pbg,
  title={{PyTorch-BigGraph: A Large-scale Graph Embedding System}},
  author={Lerer, Adam and Wu, Ledell and Shen, Jiajun and Lacroix, Timothee and Wehrstedt, Luca and Bose, Abhijit and Peysakhovich, Alex},
  booktitle={Proceedings of the 2nd SysML Conference},
  year={2019},
  address={Palo Alto, CA, USA}
}