Most knowledge graphs are incomplete. Inference of missing knowledge during the query time is important in many applications. The student will focus on pretrained neural networks to answer queries over incomplete knowledge graphs. To this end, the student will closely work on ontolearn and dice-embeddings.
A working simple example:
Graph={("Scientist","subclass","Person"),("a","type","Scientist")}
tp=Triplestore(G)
ntp=NeuralTriplestore(G)
assert len(tp.query(("?x","type","Person")))==0
assert ntp.query(("?x","type","Person"))=={"a"}
In case you have further questions, feel free to contact Caglar Demir.