Hi Clive,
PSI wouldn't categorize Rex as a concept automatically, but it
wouldn't be too hard to add a layer to draw this inference. One would
expect the semantic vectors for Fido and Rex to be similar, so finding
the nearest neighboring semantic vector, Sn to Rex and searching Sn
ISA should work (assuming we search through the elemental vectors).
It should also be possible to do this with the superposition of the
set of nearest neighboring semantic vectors to Rex.
Another way to go about this might be to superpose the nearest
neighboring semantic vectors returned by the search ISA dog (the
elemental vector), and see how far Rex falls from these. It's
interesting to think about these options as they relate to cognitive
theories of categorization.
Regards,
Trevor