Standard ordering of IPA characters?

[alice]

Is there actually such a thing, which (to pick a totally random example of the top of my head) one might pass an argument to a string comparison function such as one would use in a sort? My researches so far seem to suggest "no", but perhaps one of you knows better.

[zompist]

Is there actually such a thing, which (to pick a totally random example of the top of my head) one might pass an argument to a string comparison function such as one would use in a sort?

Maybe Unicode order?

But if you're dealing with a particular language, it should be that language's order. E.g. if you're sorting Ewe words, you want to put your Ɛ after E and your Ŋ after N.

[elgis]

If you take the unicode order route, make sure to normalize the strings before comparing them. You'll have to consider things like v̥ vs. v̊ and t̠ʃ vs t̠͡ʃ vs. t̠͜ʃ.

I don't know if this is standard, but you can maybe look at what PHOIBLE does: ordering of diacritics and modifier letters.

[WeepingElf]

Is there actually such a thing, which (to pick a totally random example of the top of my head) one might pass an argument to a string comparison function such as one would use in a sort?

Maybe Unicode order?

Hardly convenient, given that the IPA letters are in different code blocks.

But if you're dealing with a particular language, it should be that language's order. E.g. if you're sorting Ewe words, you want to put your Ɛ after E and your Ŋ after N.

This is not really IPA, but merely an extension of the Latin alphabet by adding letters derived from IPA letters. It surprises me that you haven't got that. Ɛ and Ŋ aren't even IPA letters - IPA has no capital letters.

I'd use an order for IPA which goes through the manners of articulation, preferably from least sonorous to most sonorous, and within each manner, from bilabial to glottal. And after the consonants, the vowels - from close to open, and from front to back.

[alice]

I was thinking in general terms, not for a particular language. Unicode order isn't helpful, as WeepingElf points out, since it would be nice to have, for example, <ɛ> after <e> and before <f>.

I know it's not that difficult to come up with such an ordering. The Phonetic Symbol Guide by Pullum and the other author whose name has slipped my mind again has a sensible one, but I don't know if it's any kind of standard. I was really wondering if someone had already done it, just in case there's a need to sort IPA representations somewhere.

*Maybe someone could ask ChatGPT or its brethren, just for fun? "Generate a sensible ordering for the entire set of IPA characters!"

[masako]

I was thinking in general terms, not for a particular language. Unicode order isn't helpful, as WeepingElf points out, since it would be nice to have, for example, <ɛ> after <e> and before <f>.

I hold the position that because it is a Latin based system, it should be ordered accordingly.

[WeepingElf]

I was thinking in general terms, not for a particular language. Unicode order isn't helpful, as WeepingElf points out, since it would be nice to have, for example, <ɛ> after <e> and before <f>.

I hold the position that because it is a Latin based system, it should be ordered accordingly.

Yes, this makes sense. Most IPA letters are variants of Latin letters, and could be ordered immediately following the basic Latin letters they are derived from. IPA letters which cannot easily be reduced to Latin letters that way, such as the glottal stop, may be placed at the end of the alphabet.

[zompist]

But if you're dealing with a particular language, it should be that language's order. E.g. if you're sorting Ewe words, you want to put your Ɛ after E and your Ŋ after N.

This is not really IPA, but merely an extension of the Latin alphabet by adding letters derived from IPA letters. It surprises me that you haven't got that. Ɛ and Ŋ aren't even IPA letters - IPA has no capital letters.

This strikes me like the classic statement about spherical cows. IPA borrowed e from the Roman alphabet, ɛ from Greek, ŋ from previous practice— Ben Franklin had a form of it. Ewe borrowed ɛ from IPA; once it did, it's fully Ewe, but it seems strange to me to want to deny that it's also IPA.

[alice]

Most IPA letters are variants of Latin letters, and could be ordered immediately following the basic Latin letters they are derived from. IPA letters which cannot easily be reduced to Latin letters that way, such as the glottal stop, may be placed at the end of the alphabet.

This is what Pullum and Laduslaw did, although I disagree with some of their decisions; I prefer to order <θ> and <ɯ> with <t> and <u>, for example, rather than with <o> and <m>. This could be the basis for a survey - what do others think?

[WeepingElf]

But if you're dealing with a particular language, it should be that language's order. E.g. if you're sorting Ewe words, you want to put your Ɛ after E and your Ŋ after N.

This is not really IPA, but merely an extension of the Latin alphabet by adding letters derived from IPA letters. It surprises me that you haven't got that. Ɛ and Ŋ aren't even IPA letters - IPA has no capital letters.

This strikes me like the classic statement about spherical cows. IPA borrowed e from the Roman alphabet, ɛ from Greek, ŋ from previous practice— Ben Franklin had a form of it. Ewe borrowed ɛ from IPA; once it did, it's fully Ewe, but it seems strange to me to want to deny that it's also IPA.

You have missed my point. Of course, ɛ and ŋ are IPA letters - but the capital letters Ɛ and Ŋ aren't! IPA as such has no capitals, even if some individual letters have capital counterparts in their source alphabets, or ones retro-fitted to them - which often clash: is B the capital of b, of β or of ʙ? Of course, you could invent fancy capitals for all IPA letters, but to which purpose? There are plenty of alphabetic scripts that get along just fine without case distinction.

But we are off on a tangent - the topic of this thread is not whether IPA has capitals or not, but what is the standard alphabetic order of IPA - a different question entirely.

[zompist]

You have missed my point. Of course, ɛ and ŋ are IPA letters - but the capital letters Ɛ and Ŋ aren't!

And I didn't say they were. I did use the capital forms when talking about sorting in Ewe, assuming they would make the point that the order of ɛ will differ by language. And that in turn could be important depending on what usages alice is trying to address. A universal sorting of IPA, like a universal sorting of all letterforms, will get particular languages wrong.

I prefer to order <θ> and <ɯ> with <t> and <u>, for example, rather than with <o> and <m>. This could be the basis for a survey - what do others think?

I agree with your intuition— Pullum & Ladislaw were presumably thinking about typological origins; ɯ is a turned m. On the other hand, I'm sure you recall xkcd's warning about standards.

FWIW the Apple character viewer has its own order, which doesn't follow Unicode's ordering at all. They agree with Pullum & Ladislaw on ɯ, placing it after m; but they put ɔ after o, not after c. But at some point they seemed to give up, creating multiple subsequences-- e.g. they place the small caps variants like ᴀᴄᴅ together, leaving out ʙɪʀ... it's a weird list.

[alice]

On the other hand, I'm sure you recall xkcd's warning about standards.

Oh yes; I remember it very well from the time when I asked about harmonising the syntax of SCAs.. But I'm not trying to create a standard here, just wondering if there already is one, or anything like it. The answer seems to be "no".

That said, it might still be interesting to hear conflicting opinions on the subject; it's a good source of deeply-held nitpicks

FWIW the Apple character viewer has its own order, which doesn't follow Unicode's ordering at all. They agree with Pullum & Ladislaw on ɯ, placing it after m; but they put ɔ after o, not after c. But at some point they seemed to give up, creating multiple subsequences-- e.g. they place the small caps variants like ᴀᴄᴅ together, leaving out ʙɪʀ... it's a weird list.

This is definitely not how to do it!

[WeepingElf]

You have missed my point. Of course, ɛ and ŋ are IPA letters - but the capital letters Ɛ and Ŋ aren't!

And I didn't say they were. I did use the capital forms when talking about sorting in Ewe, assuming they would make the point that the order of ɛ will differ by language. And that in turn could be important depending on what usages alice is trying to address. A universal sorting of IPA, like a universal sorting of all letterforms, will get particular languages wrong.

OK, so we misunderstood each other. I apologize.

I prefer to order <θ> and <ɯ> with <t> and <u>, for example, rather than with <o> and <m>. This could be the basis for a survey - what do others think?

I agree with your intuition— Pullum & Ladislaw were presumably thinking about typological origins; ɯ is a turned m. On the other hand, I'm sure you recall xkcd's warning about standards.

FWIW the Apple character viewer has its own order, which doesn't follow Unicode's ordering at all. They agree with Pullum & Ladislaw on ɯ, placing it after m; but they put ɔ after o, not after c. But at some point they seemed to give up, creating multiple subsequences-- e.g. they place the small caps variants like ᴀᴄᴅ together, leaving out ʙɪʀ... it's a weird list.

That's just inconsistent! I think non-Latin IPA letters, if we use the Latin alphabetic order as a starting point, should be sorted by phonetic rather than typographic considerations, so ɯ comes after u, ɔ after o, and θ after t.

[Travis B.]

That's just inconsistent! I think non-Latin IPA letters, if we use the Latin alphabetic order as a starting point, should be sorted by phonetic rather than typographic considerations, so ɯ comes after u, ɔ after o, and θ after t.

I am with you and the others here, most definitely.

[Richard W]

If you take the unicode order route, make sure to normalize the strings before comparing them. You'll have to consider things like v̥ vs. v̊ and t̠ʃ vs t̠͡ʃ vs. t̠͜ʃ.

I don't know if this is standard, but you can maybe look at what PHOIBLE does: ordering of diacritics and modifier letters.

What do you mean by 'Unicode order'? I would suggest that you take the usual default orders - I think the DUCET (Default Unicode Collation Element Table) and CLDR default orders agree for IPA letters and diacritics. As to normalisation, you convert to form NFD before applying the raw tables. 'Obviously' one wouldn't use the code point orders - that is WRONG for human-intelligible sorting.

[elgis]

I actually just meant the code point orders. I didn't know there was a non-locale-specific collation order for unicode.