From X-SAMPA phonology to screenreader-friendly script

[finlay]

Thai:
Tones:
mid 33
low 21 or 11
falling 51
high 45 or 55
rising 114 or 14
although wikipedia has this slightly cryptic aside (great example why i wish people would just use proper terminology):

Note that the full complement of five tones exist only on live syllables, those that end in a long vowel or a sonorant (/m/, /n/, /N/, /j/, /w/). For dead syllables, which end in either a plosive (/p/, /t/, or /k/) or a short vowel, only two tonal distinctions are possible: Falling vs. low for syllables containing a long vowel, and high vs. low for syllables containing a short vowel (i.e. ending either in a short vowel + plosive, or in a short vowel alone). Note that syllables that phonemically end in a short vowel have an automatic glottal stop added to the end (especially in slower speech). Hence, an alternative characterization of dead syllables is simply those syllables ending in a phonetic plosive consonant.

Consonants:
Initials:
m n N p p_h b t t_h d k k_h ? f s h ts\ ts\_h r j w l
Finals:
m n N p t k j w ? (glottal stop appears after a short vowel in the absence of another consonant)

There are only a limited number of consonant clusters:
kr kl kw k_hr k_hl k_hw pr pl p_hr p_hl tr
and a few more from loanwords like fr t_hr

Vowels:
all can appear long or short:
i M u
e 7 o
E a O

diphthongs are:
a:j aj a:w aw i:a ia iw u:a ua u:j uj e:w ew Ew M:a Ma 7:j O:j o:j

triphthongs are:
iaw uaj Maj

[finlay]

ok tired now let us know when you need more.

[clawgrip]

I wouldn't be surprised if "live" and "dead" are the native Thai terms for those syllables.

[Trebor]

ok tired now let us know when you need more.

Haha. Great, thanks so much for the info you've provided so far. I'd like to submit this assignment by next class on Tuesday or, at the latest, the following one this Thursday.

Edit: Have you looked at Omniglot's language-specific pages? They may handle phonemic inventories with more clarity, which could be helpful for, e.g., Malayalam. (They didn't look accessible in this regard, but someone sent me phonemic inventories from Omniglot a long time ago.)

[finlay]

Omniglot has a focus on displaying the script of a language. It's possible to use it to deduce the phonemic inventory, but typically wikipedia has more complete information, and will usually organize phonemes into a table based on position and manner.

That said one advantage is that it tends to include the declaration of human rights as a sample text in each language.

[Trebor]

Thai:
Tones:
mid 33
low 21 or 11
falling 51
high 45 or 55
rising 114 or 14
although wikipedia has this slightly cryptic aside (great example why i wish people would just use proper terminology):

Note that the full complement of five tones exist only on live syllables, those that end in a long vowel or a sonorant (/m/, /n/, /N/, /j/, /w/). For dead syllables, which end in either a plosive (/p/, /t/, or /k/) or a short vowel, only two tonal distinctions are possible: Falling vs. low for syllables containing a long vowel, and high vs. low for syllables containing a short vowel (i.e. ending either in a short vowel + plosive, or in a short vowel alone). Note that syllables that phonemically end in a short vowel have an automatic glottal stop added to the end (especially in slower speech). Hence, an alternative characterization of dead syllables is simply those syllables ending in a phonetic plosive consonant.

This is indeed confusing. It seems that only long vowels can take all five of the tones.

If that conclusion is correct, I could use doubled vowels to indicate vowel length and reduce the number of diacritics needed: e.g., /a:_M/ <aa>, /a:_H/ <áá>, /a:_L/ <àà>, /a:_F/ <áà>, /a:_R/ <àá>.

If anyone reading knows something about Thai phonology, please chime in.

[finlay]

It says that five tones can occur on a syllable ending in a long vowel or a sonorant. Thus short vowels in a syllable ending in a sonorant can have five tones, while long vowels in a syllable ending with an obstruent can only have the two.

So,
Long vowel + ptk: falling, low
Short vowel + ptk #: high, low
Long vowel +#, any vowel + mnNjw: mid, high, low, falling, rising

I transcribed wikipedia's table which notes the relative dearth of possible coda consonants. It mainly functions to show the mergers which take place in the script. The script is kinda complicated as it is, with quite a lot of letters standing for the same consonant but different tone, and then they're just merged further in the coda.

[Trebor]

It says that five tones can occur on a syllable ending in a long vowel or a sonorant. Thus short vowels in a syllable ending in a sonorant can have five tones, while long vowels in a syllable ending with an obstruent can only have the two.

So,
Long vowel + ptk: falling, low
Short vowel + ptk #: high, low
Long vowel +#, any vowel + mnNjw: mid, high, low, falling, rising

I transcribed wikipedia's table which notes the relative dearth of possible coda consonants. It mainly functions to show the mergers which take place in the script. The script is kinda complicated as it is, with quite a lot of letters standing for the same consonant but different tone, and then they're just merged further in the coda.

Ahh. Thanks for clarifying. I thought I may have overlooked something.

Well, it looks as though Thai is virtually impossible to transcribe in the Latin alphabet without use of less-common (and thus Jaws-unreadable) diacritics.

One possible solution comes to mind: using numbers after each syllable to indicate the tone. Does/can the average sighted person process enough letters at once to be able to obtain this information before reading a word aloud?

[R.Rusanov]

Generalizing about "sighted" people is just as stupid as generalizing about the blind. Stupider, even, given that we make up a much larger sector of the population and have correspondingly more diversity. Not all "sighted" people's brains function the same way and thinking they do is racist

[Click]

Generalizing about "sighted" people is just as stupid as generalizing about the blind. Stupider, even, given that we make up a much larger sector of the population and have correspondingly more diversity. Not all "sighted" people's brains function the same way and thinking they do is racist

?

[Jipí]

Generalizing about "sighted" people is just as stupid as generalizing about the blind. Stupider, even, given that we make up a much larger sector of the population and have correspondingly more diversity. Not all "sighted" people's brains function the same way and thinking they do is racist

Says the one who told people to just ignore that one blind person, because pff one disabled person, c'mon sheeple, he ain't worth it.

[finlay]

It says that five tones can occur on a syllable ending in a long vowel or a sonorant. Thus short vowels in a syllable ending in a sonorant can have five tones, while long vowels in a syllable ending with an obstruent can only have the two.

So,
Long vowel + ptk: falling, low
Short vowel + ptk #: high, low
Long vowel +#, any vowel + mnNjw: mid, high, low, falling, rising

I transcribed wikipedia's table which notes the relative dearth of possible coda consonants. It mainly functions to show the mergers which take place in the script. The script is kinda complicated as it is, with quite a lot of letters standing for the same consonant but different tone, and then they're just merged further in the coda.

Ahh. Thanks for clarifying. I thought I may have overlooked something.

Well, it looks as though Thai is virtually impossible to transcribe in the Latin alphabet without use of less-common (and thus Jaws-unreadable) diacritics.

One possible solution comes to mind: using numbers after each syllable to indicate the tone. Does/can the average sighted person process enough letters at once to be able to obtain this information before reading a word aloud?

Personally I don't like it, but a lot of people write pinyin in this way. It's possible definitely. And yeah people are different - i think the average person processes a whole word at once rather than letter by letter (proof of this can be found in for instance a famous experiment where you mix up the letters in a word but keep the starting and ending letters the same - and most people will be able to understand it. As i understand it, people who process letter by letter are probably labeled dyslexic.

By the way do you have a list of diacritics that you can use?

Also with regard to Thai, iirc the only offical romanization doesn't mark tone at all, so there is that option. The other is to do a Hmong and use extra coda consonants to stand for tone (Hmong used to be written Hmoob, where the double o means nasalization and the b marks a tone, since it is otherwise an illegal coda consonant. There's at least one system for Chinese that does the same.)

[Trebor]

Personally I don't like it, but a lot of people write pinyin in this way. It's possible definitely. And yeah people are different - i think the average person processes a whole word at once rather than letter by letter (proof of this can be found in for instance a famous experiment where you mix up the letters in a word but keep the starting and ending letters the same - and most people will be able to understand it. As i understand it, people who process letter by letter are probably labeled dyslexic.

By the way do you have a list of diacritics that you can use?

Also with regard to Thai, iirc the only offical romanization doesn't mark tone at all, so there is that option. The other is to do a Hmong and use extra coda consonants to stand for tone (Hmong used to be written Hmoob, where the double o means nasalization and the b marks a tone, since it is otherwise an illegal coda consonant. There's at least one system for Chinese that does the same.)

I don't like the idea of sprinkling numbers in the midst of letters either. It's useful to know that most people read word by word and not letter by letter, though. I would likely have gone for the approach formerly taken by Hmong (I couldn't justify failing to represent tone to myself), but for the difficulty that Thai allows some coda consonants: /pa_Ft/ as <pabt> and /t_h7_R:k/ as <thôôqk> don't seem very elegant at all.

The accented vowels accessible to me are as follows: <a e i o u y> with acute; <a e i o u> with grave, circumflex, and trema; <a o> with tilde; and the <ae> and <oe> ligatures.

Also, if you could X-Sampa-ize the phonemic inventories for Lao, Akan, Ewe, and Mazahua, that would be great.

[Trebor]

Generalizing about "sighted" people is just as stupid as generalizing about the blind. Stupider, even, given that we make up a much larger sector of the population and have correspondingly more diversity. Not all "sighted" people's brains function the same way and thinking they do is racist

Studies referred to by Finlay clearly had authors who disagreed, reaching conclusions the nature of which indicates that one can, in fact, make generalizations about how sighted people process text. He, for one, understood what I was driving at. It would be greatly appreciated if you learned to read others' comments in such a fashion that your goal is to enlighten the author, not diminish him.

[Ser]

1. Lao:

/b d/
/p t ts\ ts\_w k k_w ? ?_w/
/p_h t_h t_w_h k_h k_w_h/
/f s s_w h/
/m n J N N_w/
/l l_w j w/

/i: i M: M u: u/
/e: e 7: 7 o: o/
/E: E O: O/
/A: A/
/i:@_^ i@_^ M:@_^ M@_^ u:@_^ u@_^/

24 or 214 (rising)
4 (high level)
53 (high falling)
3 (mid level)
1 (low level)
31 (low falling)

2. Ewe:

/p t k kp)/
/m or b ; d ; n or d` ; J or j ; N or g ; gb)/
/ts)/
/p\ f_d s x/
/B ; v_d ; z ; G or w ; R or h\/
/l or l~/

/i i~ u u~/
/e e~ o o~/
/E E~ O O~/
/a a~/

The language is tonal, but there are contradicting accounts of tone in Ewe freely accessible on the Internet, by amateurs, often confusing a featural analysis of tone combining H and L (high and non-high), tonemes proper (phonologically distinctive tones), and the phonetic realizations of such tonemes. I don't really feel like devoting time to understand this by consulting real academic sources though...

3. Akan:

/p t k k_w/
/b d g g_w/
/f s h h_w/
/m n n_w/
/n: n:_w/
/r w/

_A stands for advanced tongue root.

/i_A i e_A e a o o_A u u_A/
(Note: the vowel qualities of the vowels with advanced tongue root are tenser than those of the others. Wikipedia claims /i/ and /e_A/, and /u/ and /o_A/, tend to get merged in pronunciation.)
/i~_A i~ a~ u~ u~_A/

Three tones: high, mid and low.

4. Mazahua:

_> after nasals and approximants stands for "glottalization"

/m m_> m_0 n n_> n_0 J J_> J_0/
/b_< d_</
/p p_h t t_> t_h k k_> k_h k_w k_w_> k_w_h ?/
/g g_w/
/ts ts_> ts_h tS tS_> tS_h/
/s s_> s_h S h/
/z Z/
/l j j_> j_0 w w_> W/
/r/

/i e @ E a O o u/
/i~ e~ E~ a~ O~ o~ u~/

[finlay]

Personally I don't like it, but a lot of people write pinyin in this way. It's possible definitely. And yeah people are different - i think the average person processes a whole word at once rather than letter by letter (proof of this can be found in for instance a famous experiment where you mix up the letters in a word but keep the starting and ending letters the same - and most people will be able to understand it. As i understand it, people who process letter by letter are probably labeled dyslexic.

By the way do you have a list of diacritics that you can use?

Also with regard to Thai, iirc the only offical romanization doesn't mark tone at all, so there is that option. The other is to do a Hmong and use extra coda consonants to stand for tone (Hmong used to be written Hmoob, where the double o means nasalization and the b marks a tone, since it is otherwise an illegal coda consonant. There's at least one system for Chinese that does the same.)

I don't like the idea of sprinkling numbers in the midst of letters either. It's useful to know that most people read word by word and not letter by letter, though. I would likely have gone for the approach formerly taken by Hmong (I couldn't justify failing to represent tone to myself), but for the difficulty that Thai allows some coda consonants: /pa_Ft/ as <pabt> and /t_h7_R:k/ as <thôôqk> don't seem very elegant at all.

The accented vowels accessible to me are as follows: <a e i o u y> with acute; <a e i o u> with grave, circumflex, and trema; <a o> with tilde; and the <ae> and <oe> ligatures.

you have enough!
mid: a (no diacritic)
rising: á
falling: à
high: â
low: ä

then you represent the other vowels with digraphs and just mark the tone on one of the letters. should be fine! (admittedly the long vowels may cause some stress)

(also i didn't really know there was a "study" tbh i've just seen the various posts on facebook illustrating it)

[Herr Dunkel]

Oh my... I... I never knew.
Rusanov can go suck a quintuple hydra dick that ableist bastard.
Wow.

Kti:

Consonants

/t d k ?/ <t d k ' >
/s z S Z x/ <s z sh zh h>
/m n r/ <m n r>

Vowels

/i i: u u:/ <i ī u ū>
/E E: 3/ 3/: O O:/ <e ē æ ǣ o ō>
/a a:/ <a ā>
(The vowels are written as letters with a macron; they're <a e i o u>, with an added ash letter - <æ> - and each with a long counterpart)

Diphthongs

/Qy uy uV ya j3\ / <ai ui ua ia iæ> (the last one's with the ash)

Triphthongs

/e}A A13\ u}3\ / <eia aie uiæ> (the last one's with the ash as well, since IDK if Jaws can read it or not and wanna be safe)

[Trebor]

Hi again everyone--I'd like to revive this thread to continue looking at how natlangs/conlangs can be given a screenreader-accessible orthography.

The first item of business: Vietnamese. I'd like to ask if anyone might be so kind as to provide its consonants, vowels, diphthongs, and triphthongs in both the current orthography and X-SAMPA. (In an alternate universe where my screenreader could read the language, I might be tempted to study it.)

Thanks.

[Salmoneus]

Hiya, Trebor!

I don't have anything to contribute to the thread, I'm afraid, but it's still good to see you around here again.

[Trebor]

Hiya, Trebor!

I don't have anything to contribute to the thread, I'm afraid, but it's still good to see you around here again.

Ahh, Salmoneus, thank you. It's good to run into you again here, too.

[Nortaneous]

I'll try to provide an inventory that marks everything the orthography does, but apparently the dialects merge some of the initials.

p t ts` c k <p t tr ch c~k>
t_h k_h~x <th kh>
b_< d_< <b đ>
f v s z s` G h <f v x d s g h>
m n J N <m n nh ng>
l r j w <l r dunno>

<d> and <gi> are both /z/. they were /D/ and /j\/ respectively in Middle Vietnamese but I don't think they're still contrasted)

Final consonants:
p t c k m n J N

Vowels:
a a: E @ O e @: o i 1 u
aj a:j @j @:j Oj oj 1j uj
aw a:w @w Ew ew 1w iw
wi
i@ 1@ u@
i@w u@j 1@w 1@j

Tones:
mid level (unmarked)
low falling breathy (grave accent)
dipping (hook)
glottalized rising (tilde)
high rising (acute)
glottalized falling (dot below)

[finlay]

Hi again everyone--I'd like to revive this thread to continue looking at how natlangs/conlangs can be given a screenreader-accessible orthography.

The first item of business: Vietnamese. I'd like to ask if anyone might be so kind as to provide its consonants, vowels, diphthongs, and triphthongs in both the current orthography and X-SAMPA. (In an alternate universe where my screenreader could read the language, I might be tempted to study it.)

Thanks.

I dunno if this will actually help, but http://www.vnspeak.com/ is an online text-to-speech tool for Vietnamese. You just need to copy and paste and click the "speak it" button and it should say it. It's not very good quality but it might give you an idea of how it is pronounced.

https://en.wikipedia.org/wiki/Vietnames ... d-Readable - This is an example of an ASCII transcription of Vietnamese which I found by googling.
https://en.wikipedia.org/wiki/Telex_%28IME%29 - This is how you type Vietnamese on an ASCII keyboard – the diacritics are replaced by digraphs and tone letters at the end of each word/syllable.

I'm going to have dinner now but maybe later I could try and tackle the task of transcribing the slightly confusing wikipedia article about the vietnamese alphabet...

[finlay]

The complete Vietnamese alphabet is:
a ă (a-breve) â (a-circumflex) b c d đ (d-bar) e ê (e-circumflex) f g h i j k l m n o ô (o-circumflex) ơ (o-horn) p q r s t u ư (u-horn) v w x y z

The telex system replaces d-bar with dd, a-breve with aw, o-horn with ow, u-horn with uw, and the three circumflex characters with double letters, although I think your system can read that so you might be able to get away with not replacing those (apparently Vietnamese orthography also contains double vowels anyway, such as cải xoong which must be input as cari xooong in the telex system).

The tone diacritics under telex are f for falling (grave), s for rising (acute), r for dipping-rising (hook), x for rising-glottalized (tilde), j for falling-glottalized (underdot)

Anyway, you could use that, or I just managed to find a converter where you can copy in Vietnamese text and get some kind of hopefully-readable script – there are a few different options to try in the menu items but Telex isn't one of them because it's primarily an input method. Also, you'd have to learn to read them. I think you'll get the best results with VIQR, which seems to replace the accents with various punctuation characters. http://www.enderminh.com/minh/vnconversions.aspx

[Nortaneous]

Vietnamese orthography contains double vowels?

[Trebor]

Thanks to Nortaneous and Finlay for your posts about the Vietnamese phonemic inventory and pointers to other ways devised for writing it in ASCII.

First, there's no need to drastically reform the orthography for the consonants, but I would like to see /f/ written as <f>, not <ph> (which is what I'd read it is currently); /d_</ as <dh>, not the special <d>; /z~j/ only as <d>, not also <gi>; /k/ only as <k>, not also <c>; and /N/ only as <ng>, not also <ngh>.

Next, I really wish the Latin alphabet had more than five basic vowel symbols, but alas, there isn't much we can do. I'm very thankful, though, to find out that Vietnamese doesn't have /M/ or /7/--I was under the impression it had one or both. So, here's my tentative proposal:

Monophthongs: /a a: E @ O e @: o i 1 u/ <a a: e er o eh er: oh i ir u> (if Mohawk can use a colon to denote vowel length even in its regular spelling system, why not Vietnamese?)
Diphthongs (group 1): /aj a:j @j @:j Oj oj 1j uj/ <ay ay: ery ery: oy ohy iry uy>
Diphthongs (group 2): /aw a:w @w Ew ew 1w iw/ <aw aw: erw ew ehw irw iw>
Diphthongs (group 3): /wi/ <wi> (no others?)
Diphthongs (group 4): /i@ 1@ u@/ <ier irer uer>
Triphthongs: /i@w u@j 1@w 1@j/ <ierw uery irerw irery>

Finally, I see no reason to scrap the present tone-marking scheme altogether. However, a few of the diacritics must be replaced, in order for (my version of) JAWS to be able to read the orthography. So, the dipping tone can be marked with a circumflex accent, not a hook; the glottalized rising tone with an acute accent followed by an apostrophe, not a tilde; and the glottalized falling tone with a grave accent followed by an apostrophe, not a dot below.

This reform clearly needs work; any comments or suggestions are of course welcome, as it's clear that the vocalic inventory looks mostly ridiculous.