Adapt the existing sequence logo to visualize base-pair interactions

The existing 'consensus' sequence logo is calculated using 'per column' symbol frequencies. This idea can be extended to pairwise sequence logos, where for an alignment like:
>sa
CCCGU
>sb
UUUGC

and secondary structure:
((.))

The logo at each position reflects the percentage of each type of base pair:

Position 1: [ CU ] 50%; [UC] 50%.
Position 2: [ UG ] 50%; [ CG ] 50%.

Position 4 and 5 would give the inverse view, ie:

Position 1: [ UC ] 50%; [ CU ] 50%.
Position 2: [ GU ] 50%; [ GC ] 50%.