Flammie Pirinen

> Well, using laplacian smoothing will solve the problem while ensuring that OOV tokens get the highest -log(P) value. Yes that should be good. > OTOH, lt-print seems to not...

I think the tropical semiring weight structures we use are only well defined in R+ including positive infinity, they may kind of work with negative values and I guess one...

These should probably just be default skippable in all languages, with possibility to override. In HFST-based language modules you can of course simulate this by freely_inserting X:eps.

> does that make the fst a lot bigger? theoretically it creates |V| new arcs and then whatever detmin might do afterwards, in practice it probably needs to be tested...

> We need the non-destructive subset of NFC. E.g., we don't want "U+212B Å ANGSTROM SIGN" normalized to "U+00C5 Å LATIN CAPITAL LETTER A WITH RING ABOVE" or the other...

> https://gist.github.com/mr-martian/11dd5c4dad3861b55054a209393c1e0c > > And here's just the unconditional replacements, since I expect that's the part we're most interested in editing. Hmm, this looks all ok to me, though I...

I like the basic stats but definetely coverage / wer too, I also think if we still assume that apertium pairs fall under these: incubator / nursery / release labels...

> But badges for things that are already part of Github's UI is superfluous. We have tags that say what release-quality a pair is at. My experience is that most...

Is the digraph problem the case where you would have both t and tz as tile? That probably is an issue for non-keyboard setup too, a language learner could often...

not sure if I understand swift code but I was just reading through it maybe this does something relevant: https://github.com/divvun/giellakbd-ios/blob/a7b5c5e64edb7e454b3e0b1f93cb19b4ecb30c55/Keyboard/Controllers/KeyboardViewController.swift#L484