IPA support

[dpk]

On https://ebensorkin.wordpress.com you mention plans to add IPA support to Merriweather.

[EbenSorkin]

I think it would be great to add. It would probably take 2 weeks+ of work.

I was not sure anyone who works with IPA would be interested.

Speculatively speaking I can imagine that it could happen in 2016 especially if I knew the ways in which it would be useful to someone.

Assuming that I did add it what styles would really be most useful?

-e.

On Oct 24, 2015, at 4:36 AM, David Kendal notifications@github.com wrote:

On https://ebensorkin.wordpress.com you mention plans to add IPA support to Merriweather.

— Reply to this email directly or view it on GitHub.

[dpk]

Most IPA is in regular-weight roman type. There are exceptions — some publications (like John C. Wells’s blog) choose to offset IPA from regular text using bold-face. And some local conventions do things like putting optional sounds in italics.

I would say the most useful (perhaps obviously) is regular, then bold, then italic. There aren’t really standard italic forms of most IPA symbols, and ‘a’ in particular is problematic in italic (the IPA requires it to be two-storey to distinguish it from ‘ɑ’, which is a different vowel sound). Elsewhere I’ve seen it suggested that an italic form be developed from the handwritten forms recommended in earlier editions of the official IPA manual, but to my knowledge nobody has actually done this yet.

[davelab6]

@dpk fascinating! Would you be interested in a kickstarter or similar crowdfunding effort to make this happen?

[dpk]

I would, but I'm not sure how many other people would.

[EbenSorkin]

This looks like 1-2 months of work. We have 8 of the 86 glyphs needed. It looks simpler than it is because while many of the glyphs are modified basic latin glyph forms, the way some of them they are changed for IPA will require a bit of study on my part to get a feeling for what makes that glyph recognized.

I think adding this would coincidentally also add full or partial support for some ( but not all) additional latin script using African languages.

[AshtarBalynestjar]

I support this. As a conlanger with ties to actual linguists, I often work with the IPA, so this would benefit me greatly.

[EbenSorkin]

I had no idea what a conlanger was before today. Cool!

I should ask you all about the existing typefaces with support for IPA. Which ones seem to get things the most consistently right and which ones are problematic. For example how do you feel about Brill? What about Andika? Gentium?

As a letter maker many of the IPA letters seem not just troublesome but also like torture because they were made without thinking about the visual logic of the Latin script. They seem deliberately and almost inevitably ugly. But if I know what feels familiar then I can at least have a better shot at making something useful!

[AshtarBalynestjar]

There are a few important issues to deal with when designing an IPA font.

First, the IPA makes extensive use of combining diacritics above and below letters, so any font that supports it must have anchor points defined on practically every glyph. Fortunately, OpenType support is widespread nowadays, so there’s no need for large amounts of precomposed glyphs anymore.

However, there is a need for an IPA stylistic set, demonstrated by the following examples:

dpk has already mentioned the issue with [a] and [ɑ] in italics. Like dpk, I would advocate for something like the cursive form given in the old IPA booklets, but that’s unconventional; most IPA fonts just have a slanted two-story ‘a’ that matches with the italic.

Furthermore, the IPA glyph for beta has a serif at the bottom, whereas the Greek beta, which shares the same codepoint, shouldn't have any serifs. The original glyph for chi has a similar problem, but the official IPA chart (which is typeset in a Times New Roman clone) shows a glyph that looks indistinguishable from the Greek chi in TNR. (If you do end up distinguishing IPA chi from Greek chi, please note that the stress on IPA chi is reversed from that on x.) There are codepoints for Latin beta and chi (U+A7B5 ꞵ, U+AB53 ꭓ), but those were only recently added to Unicode; most IPA text uses the characters in the Basic Greek block.

I honestly don’t have much of an issue with Brill or Gentium. (I haven’t tried out Andika.) They are significantly more calligraphic than Merriweather, so some decisions taken by them wouldn’t apply here, like the backwards-facing U+0283 ʃ in Gentium.

As far as stylistic decisions go, I’d prefer for U+0265 ɥ and U+0270 ɰ to have a full serif on the descender, like ‘q’. (Brill, Gentium, and Cambria all have just the rightwards-pointing serif.)

(And you’re right in that some IPA letters look like they’re badly constructed. To me, the worst offender is U+0264 ɤ, which looks non-Latin no matter how you build it. The right-facing hook on U+025A ɚ and U+025D ɝ is also a problem.)

[AshtarBalynestjar]

I’m willing to comment on test sheets once this is taking off.

[AshtarBalynestjar]

From the ‘Numerals’ thread:

Also, what do you know about support for African languages? There is some overlap with IPA.

Many African languages have orthographies based on the IPA, but I don’t know how common each letter appears to be. I’ll look into it more and try to give you a list of characters that will support the most common languages.