IPA support

[dpk]

I see mentioned in #42 that you intend to add African language support to the fonts. Could you also take a look at what other glyphs you need to add to finish IPA support? The IPA is mostly Latin alphabet + African languages glyphs + rotated/flipped versions of those, + a small number of other glyphs.

Looking at the existing glyph complement of the fonts, I don't think it would be too hard to finish IPA support once African languages are added.

[pauldhunt]

Yes, hopefully we can do both African language support and IPA support at the same time. IPA seems like it can be a bit tricky as I am unsure at how well Unicode follows the most current standard. I don’t want to be using deprecated, unused characters.

[dpk]

The IPA was last revised in 2005 and Unicode is current as of that. In cases where the Association has changed the letter used for a phoneme at some point (as with the 1989 revision of click consonant characters), Unicode encodes both the old and the new glyph at different codepoints.

[pauldhunt]

The thing is, I don’t really want to include any deprecated symbols that are no longer used. Is there any reference that specifies which symbols are currently in use to the exclusion of all others?

[dpk]

I would strongly encourage you to include deprecated symbols as well as current ones. Deprecated symbols are still useful when referring to older phonetic sources that used them; also, some phoneticians prefer to keep using the older symbols (I personally dislike the new click consonant glyphs, as do some others.). For instance, the symbol ʞ was withdrawn by the IPA before it was even used because it represented an articulation which is now known to be impossible — yet we still sometimes need the symbol, to explain the context for its withdrawal from the Alphabet.

If you’re dead set on only including current symbols, the Handbook of the International Phonetic Association (1999) should be a good enough guide, though note the ⱱ symbol was added in 2005. There have been no other changes since 1999, though ᴀ was proposed and rejected in 2011. There are also one or two symbols in common use which aren’t officially standardized by the IPA, such as ᵻ and ᵿ, which are merely the standard characters ɪ and ʊ with an overline. See this list of obsolete and nonstandard symbols.

The appendix of the TIPA manual (for the TeX typesetting system) is also pretty comprehensive.

[pauldhunt]

Thanks for making the case for deprecated symbols, this gives me better insight into how these symbols would be used. BTW, the upright styles were updated today and include ‘a good chunk’ of the IPA symbols. The support is somewhat spotty, but it would be good to have some of your feedback on what is currently there so I know what issues to address when shooting for complete support.

[moyogo]

I agree with @dpk, supporting only the official IPA doesn’t reflect the actual usage. Many phoneticians use symbols slightly different from the current official IPA due to tradition and habit, like the Asian phonetical notation which is a variant of IPA and uses deprecated IPA symbols, or the Americanists phonetical notation which is more a sibling of IPA sharing a lot of symbols but also with a few differences.

However there are also completely different phonetic alphabets that are now supported by recent versions of Unicode, like the Uralic phonetic alphabet (mixed with IPA or non standard IPA symbols in the Phonetic Extensions block 1D00..1D7F), non standard IPA (in the Phonetic Extensions Supplement block 1D80..1DBF) or, since Unicode 7.0, the German dialectology phonetic notations (in the Latin Extended-E block AB30...AB6F mixed with a couple of Americanists symbol and one orthography symbol — which should also be the glyph for IPA chi—) which have a somewhat smaller niche user base.

TIPA has a few non standard or deprecated IPA symbols, including some that are not supported by Unicode. Otherwise it’s a good starting point.

The International Phonetic Association’s page for Handbook of the International Phonetic Association has a link to its official IPA Chart. They also have an old Unicode IPA Chart with a few missing recent symbols. That would be the strict minimum. Most are in the IPA Extensions block, but the new ones are in Latin Extended-* or Phonetic Extensions blocks. The obsolete and nonstandard symbols list mentioned above is a good place to start.

[dpk]

Some comments on the IPA support so far:

• The characters are very well-designed. I particularly like the shape of the loop on ʝ.
• The most significant omission from the current batch of characters is Dark L (ɫ). Apart from that, it seems to me that Source Sans Pro already has all the characters needed for transcribing most European languages.
• The bullseye/bilabial click character (ʘ) is way too narrow. It should be more like an O with a dot in the middle than a theta with the crossbar replaced by a dot (indeed, the bolder weights risk being confused with theta at present).

Excellent work so far. I can’t wait to see the rest of the IPA glyph set completed.

[pauldhunt]

@dpk Is the confusion with theta the only reason to make the bilabial click character wider, or are there other reasons? My reasoning for keeping it narrow is that if there is only a unicase version of the glyph, I have been treating these as lowercase letters to prevent them from drawing too much attention to themselves in running text.

[dpk]

Well, a wider character is also likely to be more recognizable to linguists and phoneticians. But apart from that, no.

[dylofish]

U+02A4 and U+02A7 seem to be missing. I consider those to be very important for Speech Pathology language transcription.

[brawer]

Here’s a PDF of Source Sans Pro version 2.020, rendering an IPA test page which I originally made for Noto (another open-source font). Consider improving the tie bars in Source Sans Pro:

Admittedly my IPA test page is rather incomplete; feel free to add more test cases.

[erniemarch]

release notes list Latin Letter Ain (U+1D25), which is in Unicode charts as the full size version. The glyph in the font is marked U+1D5C, which according to the charts is the superior form (which is not in the font). Please check the unicode for this glyph.

Also, tesh digraph (U+02A7) has no outline or width in romans.

[pauldhunt]

@erniemarch The unicode value listed in the release notes is wrong, it should be LATIN LETTER AIN (U+1D5C). I’ll fix the notes.

[erniemarch]

Reviewed in Roman 2.036, Italic 1.086. Tesh digraph (U+02A7) now correct in romans.

[miguelsousa]

Fixed in version 2.040.