Why is it wrong to say that a morphological analyzer outputs an association between surface forms and canonical representations? Don't I get the association surface form -> canonical representation from running it?
Well in analysis, the input is the surface form and the output is the canonical form. See also:
+ lecture slide 10/24
+ review slide 13/24 and 16/24
+ lecture slide 10/24
+ review slide 13/24 and 16/24