Generate embeddings for all entities in a graph

[HeikoPaulheim]

Maybe I'm too blind to see it, but is there a straightforward way to create embeddings for all entities in a graph? The entities argument in fit_transform is mandatory, and the _entities field in KG seems not to give me access to URIs of entities that have a label...

[GillesVandewiele]

No indeed, this is not directly supported indeed, but would be a good addition IMO (although this will typically take a lot of time to generate)

What does the _entities attribute return? I think it should be a list of Vertex objects of which you can retrieve the name to get the URI...

[HeikoPaulheim]

The name seems to contain the rdfs:label if there's any, and the URI only in case there's no label. An additional uri field in Vertex would already help doing the trick.

[GillesVandewiele]

Sorry for the late response! Agreed that this would be useful. Strange that name does not contain the URL by default however.

                for subj, pred, obj in rdflib.Graph().parse(
                    self.location, format=self.fmt
                ):
                    subj = Vertex(str(subj))
                    obj = Vertex(str(obj))

Is what it should do when you create a KG from disk. Not sure if rdflib has change but str() of a URIRef should normally return its URI?

Could you perhaps provide a minimal example where this issue occurs? I'll take a closer look in the nearby future to it.