fgullo/kger

Playing with embedding and rule mining in knowledge graphs

Knowledge Graphs - Embeddings & Rules (KGE&R)

Playing with embedding and rule mining in knowledge graphs

Data - KGs

• AMIE
  • e.g., YAGO2 sample
• KGDatasets (used, among others, in PyKEEN)

Data - Rules

• AMIE
• AMIE (from KG-Completion Re-evaluation)

Frameworks

• AMIE [paper] [github] [web]
• PyKEEN [paper] [github] [web]
• KG-Completion Re-evaluation [paper] [github] [web]
• OpenKE [paper] [github] [web]
• LibKGE [paper] [github]
• DGL-KE [paper] [github]

Training embeddings

• Done with OpenKE
• Install OpenKE by following its installation guide (it requires PyTorch, among others)
• See the Data section of the OpenKE guide for the training-data format, and see src/generate_openke_data.py to generate it
• See src/openke/train_transe_yago2sample.py for an example on how to train the well-known TransE embedding model