you could maybe export to TMX, import into SDLX, reverse source and target in TM Maintain and then export to itd, were you can count with File Statistics.
You can make an estimation of repetitions by importing the itd file you just created overwriting existing segments into a new memory and exporting again to a second itd and count the differences.
I'm not sure about how you can get any fuzzies. To get a fuzzy you need an existing memory (not this one).
Inside a memory there are no fuzzies, all segments are 100% matches otherwise it is not a memory, its a bomb.
I hope this helps.
Thks Harry and Tectranslator.
This way could be good if the sentences were not so "heavy tagged" (XML files).
And for the purpose I am following not enough information. I'd like to see how many repetitions and fuzzies are inside too.
So, it is possible to convert a TM to a "uncleaned file"?
So I could analyse this file in a empty TM, I have the DTP settings used for this XMLs, so the tags would be not a problem for the analysis.