Graph normalization currently adds to the graph relations we are mostly interested with and consider as canonical, i.e. we privilege hypernym overy hyponym, so if there is an hyponym it adds an inverse hypernym .
There are other things we could do:
check for loops in hypernyms
remove non-canonical relations
remove self-loops for reflexive relations (i.e. node similarTo itself)
Graph normalization currently adds to the graph relations we are mostly interested with and consider as canonical, i.e. we privilege hypernym overy hyponym, so if there is an hyponym it adds an inverse hypernym .
There are other things we could do:
similarToitself)