Really nice.
In fact it would be ****great**** if we can plug the comparison between the x and y cells (which can be different).
Because like that we get a simple matrix based comparator similar to the one of early paper of prejinshight work.

This is the case actually.

Consider this rather simple and small example:

SimilarityMatrix new
on: #('hello world' 'welcome to our world' 'hello means bonjour' 'bye bye')
by: [ :m1 : m2 |
(m1 substrings intersection: m2 substrings) size
/ ((1 max: (m1 substrings union: m2 substrings) size) + 1) "NB penalty for small methods"
];
gridView

The output is

It visually says that �hello world� is close to �welcome to out world� and to �hello means bonjour�, because �hello� and �world� are in common.

Alexandre