Romanization game

[zyma]

/m n ŋ/ <m n ŋ>
/p pʰ k kʰ t tʰ/ <b p g k d t>
/ѵ ɾ/ <v r>
/ɹ j ʟ/ <ṙ j ł>
/l/ <l>

/a1 o1 u1/ <á ó ú>
/a2 o2 u2/ <à ò ù>
/a3 o3 u3/ <â ô û>
/a4 o4 u4/ <ä ö ü>
/a5 o5/ <ã õ>
/a6/ <å>

/au uo ou/ <au uo ou>


Next:

/p b pʲ bʲ t d ɖ ɟ k g kʷ gʷ/
/f v θ ð s z ʃ ʒ ɕ ʑ ʜ/
/t͡s t͡ʃ t͡ɕ c͡ç/
/m mʲ n nʷ ɲ ŋ/
/j w/
/r/
/t͡ɬ d͡ɮ/
/l ʎ/

/i iː y e eː ø æ/
/a/
/u o/

[jal]

As always, it rather depends on the phonotactics and possible combinations, but as-is, I'd say something along the lines of (without any accents or the like):

/p b pʲ bʲ t d ɖ ɟ k g kʷ gʷ/
<p b py by t d dd gy k g kw gw>

/f v θ ð s z ʃ ʒ ɕ ʑ ʜ/
<f v sf zv s z sh zh sy zy h> (or perhaps <si zi> i/o <sh zh>, or even <ś ź>)
/t͡s t͡ʃ t͡ɕ c͡ç/
<ts tsh tsy ksy>

/m mʲ n nʷ ɲ ŋ/
<m my n nw ny ng>

/j w/
<y w> (much depends on whether or not there's phonemic difference between e.g. /pj/ and /pʲ/, and whether these minimal pairs need to be properly distinguished; if so, perhaps <yy ww>)
/r/
<r>
/t͡ɬ d͡ɮ/
<tl dl>

/l ʎ/
<l ly>

For the consonants, anything palataly gets <y>, anything labialy gets <w>. <dd> is a bit of an odd one out, but since it's the only retroflex, I think I can get away with it.

/i iː y e eː ø æ/
<i ii iu e ee eu ae> (again, do we need to disnguish minimal pairs here?)

/a/
<a>

/u o/
<u o>

For the vowels, double vowels to indicate long vowels, and <u> to indicate roundness. <ae> again an odd one out, but whatever.


JAL

[zyma]

(without any accents or the like)

Not a fan of diacritics?

much depends on whether or not there's phonemic difference between e.g. /pj/ and /pʲ/, and whether these minimal pairs need to be properly distinguished; if so, perhaps <yy ww>)

They're not distinguished.

/i iː y e eː ø æ/
<i ii iu e ee eu ae> (again, do we need to disnguish minimal pairs here?)

<eu ae> do conflict with diphthongs, which I didn't display on purpose.

But it doesn't really matter because this language already has an orthography. This thread is just a little game where we can practice thinking up ways to represent different inventories orthographically. Although I suppose it could be used by someone who has an inventory that they're not sure how to represent.

Also, it's customary to continue the game by leaving an inventory yourself for the next person to romanize after you're done romanizing the previous inventory.

---

Next:

/p b t d k g/
/ᵐp ᵐb ⁿt ⁿd ᵑk ᵑg/
/s/
/m n ŋ/
/ɹ/
/l/

/i e ɛ œ æ/
/u o ʌ ɔ ɒ/

[Dezinaa]

/p b t d k g/ <p b t d k g>
/ᵐp ᵐb ⁿt ⁿd ᵑk ᵑg/ <np nb nt nd nk ng>
/s/ <s>
/m n ŋ/ <m n ŋ>
/ɹ/ <r>
/l/ <l>

/i e ɛ œ æ/ <i e è ø æ>
/u o ʌ ɔ ɒ/ <u o ë ò a>

Next:

/m n/
/p t ʈ k/
/pʰ tʰ ʈʰ kʰ/
/t͡ʂ t͡ɕ t͡ɕʰ/
/s z ʂ ɕ/
/ɻ~ʐ j w ɰ/
/i u ɯ/
/e o/
/a/

[zyma]

/m n/ <m n>
/p t ʈ k/ <p t ṭ k>
/pʰ tʰ ʈʰ kʰ/ <ph th ṭh kh>
/t͡ʂ t͡ɕ t͡ɕʰ/ <ċ c ch>
/s z ʂ ɕ/ <s z ṣ x>
/ɻ~ʐ j w ɰ/ <r j w g>

/i u ɯ/ <i u y>
/e o/ <e o>
/a/ <a>



Next:

/p b t d tˤ dˤ kʲ gʲ k g ʔ/
/f θ ð s z θˤ~sˤ ðˤ~zˤ ʃ ʒ ʃˤ ʒˤ xʲ x/
/t͡s d͡z t͡ʃ d͡ʒ/
/m n ŋʲ ŋ/
/β̞ j ɰʲ ɰ/
/ɾ/

/i iː e eː/
/a aː/
/u uː o oː/

[Dezinaa]

/p b t d tˤ dˤ kʲ gʲ k g ʔ/ <p b t d tq dq ky gy k g x>
/f θ ð s z θˤ~sˤ ðˤ~zˤ ʃ ʒ ʃˤ ʒˤ xʲ x/ <f th dh s z sq zq c j cq jq hy h>
/t͡s d͡z t͡ʃ d͡ʒ/ <ts dz tc dj>
/m n ŋʲ ŋ/ <m n ñy ñ>
/β̞ j ɰʲ ɰ/ <v y wy w>
/ɾ/ <r>

/i iː e eː/ <i ī e ē>
/a aː/ <a ā>
/u uː o oː/ <u ū o ō>

Next:

/m n̪ n/
/p pʰ pʼ t̪ t̪ʰ t̪ʼ t tʰ tʼ k kʰ kʼ ʔ/
/f θ s x h/
/j/
/i e o a/

[zyma]

/m n̪ n/ <m ñ n>
/p pʰ pʼ t̪ t̪ʰ t̪ʼ t tʰ tʼ k kʰ kʼ ʔ/ <p ph pp c ch cc t th tt k kh kk q>
/f θ s x h/ <f z s x h>
/j/ <j>

/i e o a/ <i e o a>




Next:

/p b t d k g/
/s ʃ ħ ʕ/
/m n ŋ/
/ð ɹ j w/
/l/

/i y e ø a/
/ə/
/ɒ/

[loglorn]

/Next:

/p b t d k g/
/s ʃ ħ ʕ/
/m n ŋ/
/ð ɹ j w/
/l/

/i y e ø a/
/ə/
/ɒ/

<p b t d k g>
<s ṣ h ḥ>
<m n ṅ>
<ð r j w>
<l>

<i y e ø a>
<ë>
<ä>

Next:
<p pʲ p' pʲ' pʰ pʲʰ b bʲ ɓ ɓʲ ⁿb ⁿbʲ b̥ b̥ʲ bʱ bʱʲ>
<t tʲ tʷ t' tʲ' tʷ' tʰ tʲʰ tʷʰ d dʲ dʷ ɗ ɗʲ ɗʷ ⁿd ⁿdʲ ⁿdʷ d̥ d̥ʲ d̥ʷ dʱ dʱʲ dʱʷ>
<k kʲ kʷ k' kʲ' kʷ' kʰ kʰʲ kʰʷ g gʲ gʷ ɠ ɠʲ ɠʷ ⁿg ⁿgʲ ⁿgʷ g̊ g̊ʲ g̊ʷ gʱ gʱʲ gʱʷ>
<m mʲ m̥ m̥ʲ n nʲ nʷ n̥ n̥ʲ n̥ʷ ŋ ŋʲ ŋʷ ŋ̊ ŋ̊ʲ ŋ̊ʷ>
<f fʲ f' fʲ' v vʲ s sʲ sʷ s' sʲ' sʷ' z zʲ zʷ x xʲ xʷ x' xʲ' xʷ' ɣ ɣʲ ɣʷ>
<ɰ j>

<i ɯ e ɤ a>

I guess this is the less realistic inventory i've ever thought about.

[zyma]

Not a big fan of di- and trigraphs, but…

/p pʲ p' pʲ' pʰ pʲʰ b bʲ ɓ ɓʲ ⁿb ⁿbʲ b̥ b̥ʲ bʱ bʱʲ/ <p pj pp ppj ph phj b bj bb bbj ñb ñbj hb hbj bh bhj>
/t tʲ tʷ t' tʲ' tʷ' tʰ tʲʰ tʷʰ d dʲ dʷ ɗ ɗʲ ɗʷ ⁿd ⁿdʲ ⁿdʷ d̥ d̥ʲ d̥ʷ dʱ dʱʲ dʱʷ/ <t tj tų tt ttj ttų th thy thų d dj dų dd ddj ddų ñd ñdj ñdų hd hdj hdų dh dhj dhų>
/k kʲ kʷ k' kʲ' kʷ' kʰ kʰʲ kʰʷ g gʲ gʷ ɠ ɠʲ ɠʷ ⁿg ⁿgʲ ⁿgʷ g̊ g̊ʲ g̊ʷ gʱ gʱʲ gʱʷ/ <k kj kų kk kkj kkų kh khj khų g gj gų gg ggj ggų ñg ñgj ñgų hg hgj hgų gh ghj ghų>
/m mʲ m̥ m̥ʲ n nʲ nʷ n̥ n̥ʲ n̥ʷ ŋ ŋʲ ŋʷ ŋ̊ ŋ̊ʲ ŋ̊ʷ/ <m mj hm hmj n nj nų hn hnj hnų ŋ ŋj ŋų hŋ hŋj hŋų>
/f fʲ f' fʲ' v vʲ s sʲ sʷ s' sʲ' sʷ' z zʲ zʷ x xʲ xʷ x' xʲ' xʷ' ɣ ɣʲ ɣʷ/ <f fj ff ffj v vj s sj sų ss ssj ssų z zj zų x xj xų xx xxj xxų q qj qų>
/ɰ j/ <w y>

/i ɯ e ɤ a/ <i u e o a>



Next:

/t k ʔ/
/s x h/
/t͡s k͡x/
/n/
/ɹ j ɰ/

/i œ/
/ə a/
/ɯ ɔ/

[eldin raigmore]

I just don't see the need for trigraphs; although I could be wrong about that, and if so I'd appreciate being corrected and shown.

Any 'lang is unlikely to have many more than around 100 distinct phonemes; almost certainly fewer than 150 unless it's not meant for humans or not meant to be naturalistic nor realistic.

Suppose your romanization has a 26-letter alphabet. Then you could have as many as 676 digraphs, assuming you count doubled letters as digraphs. That's 702 "graphemes"(?) that could represent phonemes.
Even if every digraph's first letter had to come from some ten-letter subset and every digraphs last letter had to come from a disjoint ten-letter subset, that's 100 possible digraphs, thus 126 possible ways of symbolizing phonemes.
Or suppose you picked two fourteen-letter subsets with only two letters in common, and every digraph's first letter had to come from one of those subsets and every digraph's second letter had to come from the other. That's 196 digraphs if you allow both of the double-letter combinations, 194 otherwise. Suppose you exclude seventy of them (arbitrarily or otherwise) and also the double-letters; there'd still be 124 non-double-letter digraphs, which together with the twenty-six letters would make 150 ways to symbolize a phoneme.

With twenty-six letters there are 17,576 possible trigraphs; I just can't imagine one would need that many ways to symbolize phonemes.

---

Of course with reasonable rules there might be many fewer trigraphs.
Maybe the first letter of a trigraph would have to come from one particular eight-letter set, the middle letter from a disjoint eight-letter set, and the last letter from yet a third eight-letter set disjoint from both of the others.
Maybe of the 8*8=64 possible combinations of first-and-second-letters only half (thirty-two) actually were allowed; something like any particular first letter could be followed by four of the second letters, and any possible second letter could be preceded by four of the first letters.
Maybe of the 8*8=64 possible combinations of second-and-last-letters only half (thirty-two) actually were allowed; something like any particular second letter could be followed by four of the last letters, and any possible last letter could be preceded by four of the second letters.
Maybe of the remaining 4*8*4=128 three-letter sequences only half (sixty-four) could be legitimate trigraphs.
Even then, you'd have 64+26+however-many-digraphs-you-had ways of symbolizing phonemes; say 26 letters + 32 digraphs + 64 trigraphs = 112 122 ways of symbolizing phonemes.

---

It might be simplest to write a generative or context-free set of production rules for how to generate "grammatical" multigraphs or polygraphs.
Or you could divide your twenty-six (or thirty-six or forty-eight or fifty-two or sixty-four or whatever) letter alphabet into sets.
One set of letters can never be used in multigraphs or polygraphs.
One set of letters can occur in multigraphs or polygraphs, but only as the first letter.
One set of letters can occur in multigraphs or polygraphs, but never as the first letter.
One set of letters can occur in multigraphs or polygraphs, but only as the last letter.
One set of letters can occur in multigraphs or polygraphs, but never as the last letter.
If you're going to allow trigraphs or bigger multigraphs, you might have a set of letters that can appear as either the first or the last letter of a multigraph but never as an interior letter;
and a set of letters that can appear as interior letters of a tri-or-longer-graph but never as first letter or last letter.

You could have further "literotactic" or "graphotactic" rules concerning pairs of consecutive letters. (Let's just assume that any such rules you have are non-trivial, that is, not consequences of the rules above.)
Maybe some pairs could never occur in multigraphs or polygraphs.
Maybe some pairs could occur in multigraphs but only as the first two letters.
Maybe some pairs could occur in multigraphs but never as the first two letters.
Maybe some pairs could occur in multigraphs but only as the last two letters.
Maybe some pairs could occur in multigraphs but never as the last two letters.
If you allow tetragrams or longer multigraphs, you might have a set of letter-pairs that can occur in multigraphs but never as the first two letters and also never as the last two letters;
and (if you allow tetragrams or longer multigraphs) you might have a set of letter-pairs that can occur in multigraphs as the first two letters and also as the last two letters but never other than as the first two or the last two.

It would actually make reading easier if there were some way to tell whether or not a given string of letters were individual letters or contained a multigraph or polygraph or were contained in a multigraph or polygraph. One way would be to pick out certain letters that couldn't occur except as part of a multigraph or polygraph.
Suppose, for instance, four letters can never be part of a multigraph; four can only be the first letter of a digraph; four can only be the last letter of a digraph; four can only be the first letter of a trigraph; four can only be the middle letter of a trigraph; and four can only be the last letter of a trigraph. That's a total of twenty-four letters, and it gives 4 + 4*4 + 4*4*4 = 4+16+64 = 84 different ways of symbolizing phonemes.
If only one could never be part of a multigraph, and the other five sets contained five letters each, the resulting twenty-six letters would give 1 + 5*5 + 5*5*5 = 1+25+125 = 151 different ways of symbolizing phonemes.

---

If the maximum-length polygraph/multigraph is digraphs, the alphabet could be divided into seven disjoint sets depending on the answers to the following three questions:

1. can the letter be used stand-alone, that is, not as part of a digraph?
2. can the letter be used as the first letter of a digraph?
3. can the letter be used as the last letter of a digraph?

There are only seven sets instead of eight because for each letter at least one of the three questions must have a positive answer.
It's possible that some of the sets are empty, and necessarily at least one of them will contain more than one letter. If either no letter has a "yes" answer for question b, or no letter has a "yes" answer for question c, then there can't be any digraphs.

For instance, suppose for two letters only question a were "yes"; for two only question b were "yes"; for two only question c were "yes"; for four both a and b (but not c) were "yes"; for four both a and c (but not b) were "yes"; for four b and c (but not a) were "yes"; and for six all three questions were answered "yes". That's twenty-four letters. 2+4+4+6=16 of them could be used "stand-alone"; 2+4+4+6=16 of them could be the first letter of a digraph; and 2+4+4+6=16 of them could be the last letter of a digraph. (Six of the digraphs would be a doubled letter. Another 30 would be divided into 15 pairs, each member of each pair being the reverse of the other member of the pair.) So you'd have 16+256 = 272 ways of symbolizing phonemes; more than twice the number of phonemes the natlang with the most phonemes has. In many cases there'd be no ambiguity as to whether a given pair of letters were a digraph, or both stand-alone, or a stand-alone followed by the first letter of a digraph, or the last letter of a digraph followed by a stand-alone, or the last letter of a digraph followed by the first letter of another digraph. But in other cases there would be ambiguity; for instance suppose a word contained a sequence of three letters for each of which all three questions were answered "yes"; the middle letter of such a sequence could be interpreted at least three different ways, as a standalone or the first letter of a digraph or the last letter of a digraph. (Considering also the possible interpretations of the other two letters there'd be eight possible interpretations the sequence, unless the rest of the word disambiguates it.)

If OTOH the maximum-length polygraph/multigraph is trigraphs, the alphabet could be divided into sixty-three disjoint sets depending on the answers to six questions:

1. can the letter be used stand-alone, that is, not as part of a digraph or trigraph?
2. can the letter be used as the first letter of a digraph?
3. can the letter be used as the last letter of a digraph?
4. can the letter be used as the first letter of a trigraph?
5. can the letter be used as the last letter of a trigraph?
6. can the letter be used as the middle letter of a trigraph?

(Again, there are only 63 sets instead of 64 because for each letter at least one of the six questions must have a positive answer.)
Necessarily most of the sets (at least 37 of them) must be empty.
Perhaps (probably?) the answers to the questions aren't logically independent of each other.
Maybe if question d is answered "yes" then question b must also be answered "yes".
Maybe if question e is answered "yes" then question c must also be answered "yes".
Maybe if question f is answered "yes" then at least one of questions b or c must be answered "yes"; or, one of questions a or b or c.
Maybe if question b is answered "yes" then at least one of questions d or f must be answered "yes".
Maybe if question c is answered "yes" then at least one of questions e or f must be answered "yes".

[zyma]

I used digraphs in that particular romanization because:

a) I was typing it a few minutes before I had to leave my house, and it's a lot quicker for me to type letters on my keyboard (without diacritics).
b) I wanted some sort of symmetry and "sensibility". I mean sure, if you want to use up every possible single character and every possible digraph that doesn't conflict with a cluster to represent your phonemes, then by all means, be my guest. But take this (probably unrealistic) inventory I just made up as an example:


/p b t d c k g q/
/f v s z x h/
/m n/
/j w/
/r/
/l/

/i y e/
/a/
/u o/

I would think it would be an obvious choice to romanize this inventory with the basic 26 Latin characters.

But now, let's take out some phonemes, namely /c q x y/, and let's add /fʲ vʲ sʲ zʲ/, leaving us with:

/p b t d k g/
/f v fʲ vʲ s z sʲ zʲ h/
/m n/
/j w/
/r/
/l/

/i e/
/a/
/u o/

You could utilize every available character before moving on to any di- or trigraphs, romanizing /fʲ vʲ sʲ zʲ/ as <c q x y>, but I think it makes more sense to move on to using multiple characters for one sound before making… uncommon choices like this. If this language makes a distinction between palatalized fricatives and sequences of fricative + /j/, then we could represent palatalization with a <y>, which we're no longer using, after <f v s z>.

Similarly, I could have romanized the inventory like this:

/p pʲ p' pʲ' pʰ pʲʰ b bʲ ɓ ɓʲ ⁿb ⁿbʲ b̥ b̥ʲ bʱ bʱʲ/ <a b c d e f g h i j k l m n o p>
/t tʲ tʷ t' tʲ' tʷ' tʰ tʲʰ tʷʰ d dʲ dʷ ɗ ɗʲ ɗʷ ⁿd ⁿdʲ ⁿdʷ d̥ d̥ʲ d̥ʷ dʱ dʱʲ dʱʷ/ <q r s t u v w x y z aa ab ac ad ae af ag ah ai aj ak al am an>
/k kʲ kʷ k' kʲ' kʷ' kʰ kʰʲ kʰʷ g gʲ gʷ ɠ ɠʲ ɠʷ ⁿg ⁿgʲ ⁿgʷ g̊ g̊ʲ g̊ʷ gʱ gʱʲ gʱʷ/ <ao ap ak ar as at au av aw ax ay az ba bb bc bd be bf bg bh bi bj bk bl>
/m mʲ m̥ m̥ʲ n nʲ nʷ n̥ n̥ʲ n̥ʷ ŋ ŋʲ ŋʷ ŋ̊ ŋ̊ʲ ŋ̊ʷ/ <bm bn bo bp bq br bs bt bu bv bw bx by bz ca cb>
/f fʲ f' fʲ' v vʲ s sʲ sʷ s' sʲ' sʷ' z zʲ zʷ x xʲ xʷ x' xʲ' xʷ' ɣ ɣʲ ɣʷ/ <cc cd ce cf cg ch ci cj ck cl cm cn co cp cq cr cs ct cu cv cw cx cy cz>
/ɰ j/ <da db>

/i ɯ e ɤ a/ <dc dd de df dg>

But that doesn't make much sense to me.


And in another unrealistic hypothetical inventory, say we have:

/p b ᵐbʲʱ ᵐbˤʱ ᵐbʷʱ ɓ t d ⁿdʲʱ ⁿdˤʱ ⁿdʷʱ ɗ k g ᵑgʲʱ ᵑgˤʱ ᵑgʷʱ ɠ q͡ʡ/
/s h/
/n/
/r/

/i/
/a/
/u/

You could either romanize the stops as:

/p b ᵐbʲʱ ᵐbˤʱ ᵐbʷʱ ɓ t d ⁿdʲʱ ⁿdˤʱ ⁿdʷʱ ɗ k g ᵑgʲʱ ᵑgˤʱ ᵑgʷʱ ɠ q͡ʡ/ <p b c e f j t d l m o v k g w x y z q>

Or as:

/p b ᵐbʲʱ ᵐbˤʱ ᵐbʷʱ ɓ t d ⁿdʲʱ ⁿdˤʱ ⁿdʷʱ ɗ k g ᵑgʲʱ ᵑgˤʱ ᵑgʷʱ ɠ q͡ʡ/ <p b nbj nbx nbw bb t d ndj ndx ndw dd k g ngj ngx ngw gg q>


I guess my point is that, while I'm not too keen on the aesthetic of trigraphs, they certainly can be helpful when trying to make romanizations of larger phonologies symmetrical and sensible, especially when romanizing phonemes that implement several characters in the IPA, such as /bʱʲ/. So technically, there really is no need for trigraphs, but you could easily make a similar claim about, say, airplanes. We can live our lives just fine without them, but they sure make things a hell of a lot easier. At least in my opinion. I'm sure you could find a way to make single characters and digraphs work just fine in terms of symmetry and sensibility, using all your groups and patters that I don't fully understand, partially because it's after midnight while I'm typing this.


Looking back over this after I've written it, I want to sincerely apologize for how I come off in my response. Although it was certainly not my intention, I sound like a big, sarcastic jerk. I feel really bad about that, especially since you obviously put a lot of time and effort into your post.

[eldin raigmore]

I want to sincerely apologize for how I come off in my response.

Thanks, but, it's not necessary.

I sound like a big, sarcastic jerk.

Not to me.
I asked someone (e.g. you) to explain why trigraphs were a better choice than somehow stopping at digraphs and/or diacritics, and you did so.
That was IMO the polite thing to do; and as far as I could tell it was done in the polite manner, too.
It's up to me to decide whether or not I'm convinced by your reasoning. I haven't decided yet but I'm even more open to the possibility you're right about all that than I was before I read your post.

I would like to ask if you have any way to be sure you've minimized (or reasonably-nearly-minimized) the ambiguity problems.

[zyma]

Not to me.

Thank you; I'm glad. I guess I'm just really paranoid about unintentionally offending people.

But, to get back to the actual topic here:

I would like to ask if you have any way to be sure you've minimized (or reasonably-nearly-minimized) the ambiguity problems.

Naturally, it all comes down to phonotactics and the size of the inventory. For the purposes of this attempted explanation, I'm essentially going to ignore the possibility of di- and trigraphs representing vowels.

If a language allows only (C)V or V(C) syllables, then I don't foresee many ambiguity problems. You could theoretically have:

/p t k/ <p t ptfsxmnr>
/f s x/ <f s x>
/m n/ <m n>
/r/ <r>

/i a/ <i a>
/u/ <u>

And there would never be any ambiguity, because with strict (C)V or V(C) phonotactics, you could never have two consonants in a row. If there was a word /kamu/ <ptfsxmnramu>, that could never be interpreted as /ptfsxmnramu/, because clusters aren't allowed.

But, let's change /k/ to <tx>, add /b d g/ <mp nt ntx>, and change the phonotactics to (C)(S)V(N), where S is any fricative, and N is any nasal. Now, <txamu> could be either /kamu/, or /txamu/. <sintxu> could either be /sigu/, /sinku/, /sidxu/, or /sintxu/. How could we fix this?

Obviously, we could change the orthography. But, if I remember correctly, Catalan (or another Iberian Romance language) uses a <·> to distinguish the palatal lateral approximant <ll> from a geminate velarized coronal lateral approximant <l·l>.

We could have <sintxu> /sigu/, <sin·txu> /sinku/, <sint·xu> /sidxu/, and <sin·t·xu> /sintxu/, but that doesn't look very good.

This isn't the best example, because its phonology could easily be romanized with single characters. Let's try something larger:

/p ʰp pʰ p' pː t ʰt tʰ t' tː k ʰk kʰ k' kː ʔ/
/f fː h/
/m n/
/j w ʕ/
/ʀ/
/ɬ/
/l/


/i a/
/u/

With (C)(C)V(C)(C) phonotactics. C can be any consonant.

We could romanize it like this:

/p ʰp pʰ p' pː t ʰt tʰ t' tː k ʰk kʰ k' kː ʔ/ <p hp ph p' pp t ht th t' tt k hk kh k' kk '>
/f fː h/ <f ff h>
/m n/ <m n>
/j w ʕ/ <y w c>
/ʀ/ <r>
/ɬ/ <tl>
/l/ <l>

But then, <mathta> could be either /mathta/, /matʰ.ta/, or /mat.ʰta/. I guess you could use accent marks, like this: <mathta>/mathta/, <mathtá> /matʰ.ta/, <máthta> /mat.ʰta/. But I would personally change the orthography, so that postaspiration is represented by <x>, and preaspiration is represented by <q>. Thus, <mathta> /mathta/, matxta /matʰ.ta/, and <matqta> /mat.ʰta/. Since <p'a> could still either be /pʔa/ or /p'a/, let's keep the apostrophe for ejectives, and let's take a page from a grammar of Hawaiian by using <ʻ> for the glottal stop. And if those look too similar, we could go the Spanish-esque route with <j> /h/ and <h> /ʔ/. <tlu> could either be /tlu/ or /ɬu/, so let's make /ɬ/ <z>. And finally, geminates. According to our phonotactics, geminates could be confused with sequences of two of the same consonant. <ffi> could either be /ffi/ or /fːi/. Since there aren't any voiced stops or fricatives, /pː tː kː fː/ <b d g v> sounds like a good idea to me.

I'm going to try to give another example with trigraphs:

/p b ᵐb ᵐbʷ ᵐbʷʱ t d ⁿd ⁿdʷ ⁿdʷʱ k g ᵑg ᵑgʷ ᵑgʷʱ ʔ ʔʷ/ <p b mb mbw mbwh t d nd ndw ndwh k g ng ngw ngwh q qw>
/f v s z x h/ <f v s z j h>
/m n/ <m n>
/j w ʕ ʢ/ <y w c x>
/r/ <r>
/l/ <l>

/i e a/ <i e a>
/u o/ <u o>

Phonotactics are C(A)(A)V(N). A is any approximant, lateral approximant, trill, or fricative. N is any nasal. Note that, in the romanization, all 26 basic characters are used.

Take the word <combwhan>. I intend for that word to be pronounced as /ʕo.ᵐbʷʱan/, but how do you know it's not supposed to be /ʕom.bʷʱan/, /ʕom.bwhan/, /ʕom.bʷhan/, /ʕo.ᵐbʷhan/, /ʕo.ᵐbwhan/, or possibly other pronunciations? In my opinion, the best way to solve ambiguity problems is to get creative and to switch things around. Since all basic characters have already been used, we have to look elsewhere to solve this problem. Since the basic syllable in this language is CV, you can't have any diphthongs or sequences of vowels, so let's use <u> to represent labialization. <xambuo> is /ʢaᵐbʷo/, not /ʢaᵐbwo/ or /ʢaᵐbuo/. If you don't mind me using diacritics here, I'd recommend using <ñ> to clear up the ambiguity between words like /xandi/ and /xaⁿdi/, both <jandi>. With the addition of <ñ>, however /xaⁿdi/ is <jañdi>, while /xandi/ is still <jandi>. I believe our last problem is aspiration/breathy voiced release. When a syllable contains an aspirated stop, the vowel nucleus of that syllable gets a circumflex <ˆ>. Therefore, /ʕo.ᵐbʷʱan/ is now romanized as <coñbuân>, instead of <combwhan>, which is now clearly /ʕom.bwhan/. Unless I've overlooked something, I've successfully gotten rid of all orthographic ambiguity in this hypothetical language, and I've retained 10 polygraphs. Also, the only non-basic characters I added were <ñ â ê î ô û>.

We now have:

/p b ᵐb ᵐbʷ ᵐbʷʱ t d ⁿd ⁿdʷ ⁿdʷʱ k g ᵑg ᵑgʷ ᵑgʷʱ ʔ ʔʷ/ <p b ñb ñbu ñbuV̑ t d ñd ñdu ñduV̑ k g ñg ñgu ñguV̑ q qu>
/f v s z x h/ <f v s z j h>
/m n/ <m n>
/j w ʕ ʢ/ <y w c x>
/r/ <r>
/l/ <l>

/i e a/ <i e a>
/u o/ <u o>


Actually, I feel the need to clarify why (and when) I prefer diacritics to polygraphs, or rather, when I don't.

Let's take the stops of the last example inventory:

/p b ᵐb ᵐbʷ ᵐbʷʱ t d ⁿd ⁿdʷ ⁿdʷʱ k g ᵑg ᵑgʷ ᵑgʷʱ ʔ ʔʷ/

We have prenasalization, labialization, and aspiration/breathy voice. Let's say < ̥ > represents labialization (because it's rounded), < ̤ > represents breathy voice (I'm almost certain it's used officially as an indicator of that), and that < ̃ > represents prenasalization (because nasal vowels are marked with it). That leaves us with:

/p b ᵐb ᵐbʷ ᵐbʷʱ t d ⁿd ⁿdʷ ⁿdʷʱ k g ᵑg ᵑgʷ ᵑgʷʱ ʔ ʔʷ/ <p b b̃ b̥̃ b̥̤̃ t d d̃ d̥̃ d̥̤̃ k g g̃ g̥̃ g̥̤̃ q q̥>

Unless you're willing to make diacritics asymmetrical (like using <ḇ ḓ ǵ> for /ᵐb ⁿd ᵑg/), which many people, myself included, are not, diacritics can get ugly, confusing, and just plain impractical. I mean, sure, you could do something like:

/b ᵐb ᵐbʷ ᵐbʷʱ d ⁿd ⁿdʷ ⁿdʷʱ g ᵑg ᵑgʷ ᵑgʷʱ/ <b ḃ ḇ ƀ d ḋ ḏ đ g ġ ǥ ḡ>

If you're willing to use diacritics that have no clear ties to the sounds they represent (For example, what about <ḃ ḋ ġ> screams "prenasalized"? For me, nothing. However, many diacritics, like <ḍ> are often associated with certain sounds, in this case a voiced retroflex stop.).

In cases like this, there are many advantages to polygraphs.

Although I can't think of many other quick and easy examples with trigraphs, I hope I've demonstrated some ways to help clear up ambiguity here. If you still have any questions that weren't answered here, please ask. The larger a phonology is, and the more complicated the phonotactics are, the harder it's going to be to clear up ambiguity, and you won't be able to do that in a way that appeals to you in every way 100% of the time. Like I said before, I agree with you that trigraphs are not necessary, but I stand by my assertion that they're very helpful in some cases.

[Nortaneous]

If you take opposition to digraphs too far, you get the Iu Mien Old Roman script.

/pʰ tʰ tsʰ tɕʰ kʰ/ <P T Z Q K>
/p t ts tɕ k/ <p t z J k>
/b d dz dʑ g/ <B D R F G>
/f s h/ <f s h>
/m n ɲ ŋ/ <m n E v>
/m̥ n̥ ɲ̥ ŋ̥/ <M N H V>
/l j w/ <l y w>
/l̥ j̥ w̥/ <L Y W>
/æ a aː ɔ e ə o i u/ <c a aa x e r o i u>

Tones: <-0 -b -q -d -j -g>

It looks like this:

zinb Daavg luvb Diaq myenb yetg zuvq kxvq Du'g Euvg waag Navq. waab eij fib Navq Revg.
ninb Bua swiq Jyenq mivb tov puv mcvj, Tauj si-naa teib puv, maaib Daub pcvb Eei twvj, zyrug tivb Fiag yem uaq a'q.
laanb kxvq Buaj laanb, "taaib maa'b! Bua nanq niab zun puaq vcvg taaib." ninb Bua lovg niab zun tiq la'pyeiq. Daub Diaq Erub Haig Eei zuv ninb Bua lovg tiq pa'naib nia.
ninb Bua cvj kxvq, "taaib maa'b! Bua Zeij nxm zivb yaag Zeij nxm fanb Taapq Lav Tauj kub vwaaig luvb, pun Bua maaib mevb taub yaag maiq zu'g pun Raanj pwavq luvb Diaq"

cf.


Zinh ndaangc lungh ndiev mienh yietc zungv gorngv nduqc nyungc waac hnangv. Waac-eix fih hnangv nzengc.
Ninh mbuo suiv jienv mingh dong bung maengx, taux Sinaa Deic-Bung, maaih ndau-baengh nyei dorngx, ziouc dingh njiec yiem wuov aqv.
Laanh gorngv mbuox laanh, "Daaih maah! Mbuo nanv nie-zun buov ngaengc daaih." Ninh mbuo longc nie-zun div la'bieiv. Ndau-ndiev nyouh haic nyei zung ninh mbuo longc div ba'naih nie.
Ninh mbuo aengx gorngv, "Daaih maah! Mbuo ceix norm zingh yaac ceix norm fanh taapv hlang taux gu'nguaaic lungh, bun mbuo maaih mengh dauh yaac maiv zuqc bun nzaanx buangv lungh nidev."

[zyma]

Well, it looks like the debate has settled down here, so, in case anyone would be interested in continuing the game, I'm going to repost the latest un-romanized inventory.

Next:

/t k ʔ/
/s x h/
/t͡s k͡x/
/n/
/ɹ j ɰ/

/i œ/
/ə a/
/ɯ ɔ/

[Nortaneous]

/t k ʔ/ <d g p>
/s x h/ <s x h>
/t͡s k͡x/ <c k>
/n/ <n>
/ɹ j ɰ/ <r y w>

/i œ/ <i ö>
/ə a/ <e a>
/ɯ ɔ/ <u o>

Next:

/p b (mb) t d (nd) ts dz (ndz) tʂ dʐ tɕ dʑ (ndʑ) k g/
/f s ʂ ɕ x (h)/
/m n ɲ/
/l r j w/
/a e ə o i u z̩ v̩/
/ã ẽ ə̃ õ ĩ ũ/
/m̩ n̩ ɲ̩ ŋ̩/
Four tones: /à ā á â/

Syllable structure is CV. Nasals are realized as prenasalized stops before oral vowels; [ndz] only appears before /z̩/. Syllabic nasals can't take an onset except [h]; [h] doesn't appear anywhere else.

[Click]

/p b (mb) t d (nd) ts dz (ndz) tʂ dʐ tɕ dʑ (ndʑ) k g/ p b m t d n ts dz n tr dr c ȷ n k g
/f s ʂ ɕ x (h)/ f s sr x h 0
/m n ɲ/ m n
/l r j w/ l r y v
/a e ə o i u z̩ v̩/ a e y o ı u ey ev
/ã ẽ ə̃ õ ĩ ũ/ an en yn on ın un
/m̩ n̩ ɲ̩ ŋ̩/ ang eng ıng ong

Four tones: /à ā á â/ à a á ā

Next:

/m n ɳ ɲ/
/p t tʂ tɕ k/
/β ð s z ʂ ʐ ɕ ʑ/
/β̞ r ɺ ɽ ɺ̢ j/

/i e a o u/
/iː eː aː oː uː/
/ɪ ɐ ʊ/

Use diacritics. Consistently.

[sangi39]

/m n ɳ ɲ/ <m n ň ń>
/p t tʂ tɕ k/ <p t č ć k>
/ð s z ʂ ʐ ɕ ʑ/ <d s z š ž ś ź>
/β̞ r ɺ ɽ ɺ̢ j/ <b r l ř ľ j>

/i e a o u/ <i e a o u>
/iː eː aː oː uː/ <ï ē ā ō ū>
/ɪ ɐ ʊ/ <ì à ù>

Next:

/m n ɲ ŋ/
/p b t d k g kʷ gʷ/
/f v θ ð s z x xʷ h/
/w r l ʎ j/

/i i: y: u u:/
/e e: ø: o o:/
/a a:/

The labiovelars have limited distribution, occurring only before non-rounded vowels.

Syllable Structure: CV (C is optional word-initially but obligatory word-internally)

Extra points for not using digraphs

[Creyeditor]

/m n ɲ ŋ/ <m n ň ṅ>
/p b t d k g kʷ gʷ/ <p b t d k g ku gu>
/f v θ ð s z x xʷ h/ <f v s z š ž x xu hu>
/w r l ʎ j/ <w r l y j>

/i i: y: u u:/ <i í ü u ú>
/e e: ø: o o:/ <e é ö o ó>
/a a:/ <a á>

Next:

/m n/
/p b t d tɬ c k q/
/s ɬ ʃ h/
/r l j/

/i ɨ u/
/e ə o/
/a/

There is also a simple two tone system.

Syllable strcuture:
(C)V(C)
Onsetless syllables are only permitted word initially. A syllable with a coda is prohibited everywhere except word-finally.

Allophony
Intervocalic obstruent voicing, underlyingly voiced stops become fricatives.
Final stop devoicing.
Vowels nasalize before and after nasal consonants.
Vowels get breathy voiced before and after /h/
/h/ is deleted word finally and word initially.
Some alveolar consonants (/n t d s l/) and some velar consonants (/k g/) become palatal ([ɲ c ɟ ʃ ʎ][c ɟ]) before and after /i,e/.

[Shemtov]

/m n <m n>
/p b t d tɬ c k q/ <p b t d tl c k q>
/s ɬ ʃ h/ <s ļ š h>
/r l j/ <r l j>

/i ɨ u/ <i ï u>
/e ə o/ <e y o>
/a/ <a>


Next:
Phonology:
Consonants:
/m mʲ n ɳ ŋ/
/p pʲ t ʈ k/
/pʰ tʰ ʈʰ kʰ/
/ɸ ɸʲ f s ʂ x h/
/l ɭ ʟ/
/β̞ ɥ ɹ~ɻ  ɰ/

Vowels:
/i ɨ u ɯ/
/e ɘ o/
/ɛ ɜ ɔ ʌ/
/a ɑ/

Diphthongs:
/ai ae aɛ au aɯ ao aɔ aʌ/
/oi oe oɛ/
/ɔi/