relData2OpenKE

Scripts to translate Knowledge bases in relData format into OpenKE format. See these projects in:

• https://github.com/phohenecker/rel-data#data-format

• https://github.com/thunlp/OpenKE

relData2OpenKE.py allows to specify the folder where the KB in rel-data format is stored and the outputPath, also the partitions for training/validation/testing which are created based only on infereable data.

The output includes the OpenKE files:

• Entity2id
• Relation2id
• Train2id
• Valid2id
• Test2id
• type_constrain.txt

relData2OpenKE basically:

• uses sample2EntitiesAndTriples.py, to deal with each sample of rel-data format, e.g. 5000.individuals, 5000.classes…
• objects of sample2EntitiesAndTriples for all samples are used to generate aux files that, among others,
  • consider class belonging as another simple relation,
  • consider global identifiers for entities/relations/classes.
• uses iDManager.py to map entity/classes names to new IDs avoding repetitions.
  • relData gives different identifiers to the same individuals/classes.
• check if there are repetitions in triples for training, or triples in training also present in validation or testing.
  • relData can have these repetitions.
• and generate the final OpenKE files.