Why are some same-shape sides not sorted into groups?

Post Reply
stephennixon
Posts: 10
Joined: 18 Jun 2021

Why are some same-shape sides not sorted into groups?

Post by stephennixon »

I'm working on a Latin + Cyrillic font with a collaborator, and we are trying out KernOn for it.

My collaborator pointed out an unexpected issue: some kerns seemed to be missing, even though a same-shape-based group would put them there.

That is, in a string like "ДЧДҶДҸДӋД," we would expect the ДЧ kern to be reflected in all nearly-identical shape variations such as ДҶ and ДҸ and ДӋ.

Have I missed a step, or done something wrong? I could imagine this being an issue of the Kb data limit I am setting, but I'll ask that as a separate question, as it is somewhat distinct.

I've read elsewhere on the forum that KO avoids shape-based grouping for some space-saving consideration, but I am confused about this... does it really add a meaningful amount of data to make a left group for related shapes such as ЧҶҸӋ?

In that other post, you pointed out that a search of Wikipedia text in Russian, Ukrainian, and Belarusian didn't turn up certain pairs, and therefore they shouldn't be kerned. But, my collaborator wondered: how sure are we that those pairs don't happen in any of of the 90 or so Cyrillic languages we are targeting? We don't want to design badly for a minority language just because a major language doesn't happen to have a certain pair.

Thanks for any insights here!
Attachments
Screen Shot 2022-08-05 at 2.13.49 PM.png
Screen Shot 2022-08-05 at 2.13.49 PM.png (105.23 KiB) Viewed 61 times
stephennixon
Posts: 10
Joined: 18 Jun 2021

Re: Why are some same-shape sides not sorted into groups?

Post by stephennixon »

I see that part of my question ("how sure are we that those pairs don't happen in any of of the 90 or so Cyrillic languages we are targeting?") is answered in https://kern-on.com/under-the-hood. KO has a list of pairs to kern, determined and prioritized based on Wikipedia in 200+ languages, about 1 million tweets, and some extra weighting added to rare scripts.
stephennixon
Posts: 10
Joined: 18 Jun 2021

Re: Why are some same-shape sides not sorted into groups?

Post by stephennixon »

I guess I'm still curious why similar-shaped glyphs aren't added into groups... it seems like it might be a nearly "free" way to get kerning for more potential pairs, even if they aren't common. I think I read this discussion in another thread, so i suppose maybe this post is just to say that I would add another request for shape-based grouping.
Post Reply