Awesome Knowledge Graph Embedding Approaches

awesome-kge.md

Awesome Knowledge Graph Embedding Approaches

This list contains repositories of libraries and approaches for knowledge graph embeddings, which are vector representations of entities and relations in a multi-relational directed labelled graph. Licensed under CC0.

Libraries

• 🆕 AmpliGraph (4 algorithms) @ https://github.com/Accenture/AmpliGraph
• Embedding framework (5 algorithms) @ https://github.com/BookmanHan/Embedding
• KB2EP collection (5 algorithms) @ https://github.com/thunlp/KB2EP
• 🆕 OpenKE (8 algorithms) @ https://github.com/thunlp/OpenKE
• 🆕 PyKEEN (10 algorithms) @ https://github.com/SmartDataAnalytics/PyKEEN/
• Python KGE (6 algorithms) @ https://github.com/mana-ysh/knowledge-graph-embeddings
• scikit-kge (3 algorithms) @ https://github.com/mnick/scikit-kge
• scikit-tensor (5 algorithms) @ https://github.com/mnick/scikit-tensor
• SME (4 algorithms) @ https://github.com/glorotxa/SME

Approaches

• ComplEx @ https://github.com/ttrouill/complex
• ComplEx-NNE-AER @ https://github.com/iieir-km/ComplEx-NNE_AER
• ConvE @ https://github.com/TimDettmers/ConvE
• EbemKG @ https://github.com/pminervini/ebemkg
• 🆕 fastText @ https://github.com/facebookresearch/fastText/tree/master/scripts/kbcompletion
• HolE @ https://github.com/mnick/holographic-embeddings
• Inferbeddings @ https://github.com/uclmr/inferbeddings
• KG2Vec @ https://github.com/AKSW/KG2Vec
• KGE-LDA @ https://github.com/yao8839836/KGE-LDA
• KR-EAR @ https://github.com/thunlp/KR-EAR
• LiteralE @ https://github.com/SmartDataAnalytics/LiteralE
• mFold @ https://github.com/v-shinc/mFoldEmbedding
• Neural-LP @ https://github.com/fanyangxyz/Neural-LP
• ProjE @ https://github.com/bxshi/ProjE
• RDF2Vec @ http://data.dws.informatik.uni-mannheim.de/rdf2vec/code/
• Resource2Vec @ https://github.com/AKSW/Resource2Vec/tree/master/resource2vec-core
• TranslatingModel @ https://github.com/ZichaoHuang/TranslatingModel
• wiki2vec (for DBpedia only) @ https://github.com/idio/wiki2vec

Hello and thanks all.

@lukostaz and @mberr: I've added your libraries to the list.

@albertmeronyo: As node2vec is an embedding algorithm for graphs with only one relation, I would leave it out of the list.

Hi,
Quite recent library from Facebook research:
https://github.com/facebookresearch/PyTorch-BigGraph

Hi,
Quite recent library from Facebook research:
https://github.com/facebookresearch/PyTorch-BigGraph

Hey,

Thanks for sharing, is there any document as a guide line on how to use PytTorch-BigGraph? i would like to use this tool for Knowledge Graph embeddings.....

Hi @mommi84, here are some suggestions:

BioNEV

https://github.com/xiangyue9607/BioNEV provides:

• 5 matrix factorization-based: Laplacian Eigenmap, SVD, Graph Factorization, HOPE, GraRep
• 3 random walk-based: DeepWalk, node2vec, struc2vec
• 3 neural network-based: LINE, SDNE, GAE

Some of them might not qualify as KGE models as you mentioned in https://gist.github.com/mommi84/07f7c044fa18aaaa7b5133230207d8d4#gistcomment-2977958, but this might be a good place for people who are interested in the biological applications to start looking around!

Edge2vec

https://github.com/RoyZhengGao/edge2vec improves node2vec to consider the statistical distribution of edge-edge transitions in second-order random walks

Metapath2vec

I'm not happy that this author has opted to share code via dropbox instead of GitHub, but it can be found at https://ericdongyx.github.io/metapath2vec/m2v.html

DGL-KE： https://github.com/awslabs/dgl-ke , including TransE, TransR, RESCAL, DistMult, ComplEx, and RotatE.

Hi,

CKB is a tool to make knowledge graph embeddings using HuggingFace models. This is useful when a knowledge graph contains text for example. DistillBert and FlauBERT are available. The source code is developed so that new HuggingFace models can be added without difficulty.

CKB works with MKB which is a more traditional tool for knowledge graph embeddings. MKB makes it easy to evaluate a model using link prediction / triplet classification.

CKB and MKB lack documentation compared to the libraries already referenced.