Thanks for your contribution!!
Mine confusion is this way,
I have two monolingual embedding file say (src) A and (tgt) B.
I have applied your transformation to get two embedding files (src') C and (tgt') D.
Now I know 'A' is the projected onto 'B' which creates C.
Why D is there?
My confusion is which pair of (source, target) embedding are in common space
Whether it is (C, D) in common or (C, B) in common.
I have tested performance on (C, D) and (C, B). Performance on both pairs are almost similar, to be precise pair (C, D) is more than pair (C, B).
Please help!!
Thanks for your contribution!! Mine confusion is this way, I have two monolingual embedding file say (src) A and (tgt) B. I have applied your transformation to get two embedding files (src') C and (tgt') D. Now I know 'A' is the projected onto 'B' which creates C. Why D is there? My confusion is which pair of (source, target) embedding are in common space Whether it is (C, D) in common or (C, B) in common. I have tested performance on (C, D) and (C, B). Performance on both pairs are almost similar, to be precise pair (C, D) is more than pair (C, B). Please help!!