ConlangDictionary 0.3 - now phonology parsing is faster

[dhok]

I work Windows XP, and although it is a magnificent program, I'd like to report a bug: the program crashes when I try adding coda clusters.

Also, can we get a diacritics inserter? Works doesn't have macrons, and it's annoying to be copy-and-pasting from the internet.

[faiuwle]

I work Windows XP, and although it is a magnificent program, I'd like to report a bug: the program crashes when I try adding coda clusters.

Can I get more specifics here? Adding 3x3x3 coda clusters works fine on my end - are you trying to add {lots of consonants} + {lots of consonants}...x5 or something, like in Qwynegold's case?

Also, can we get a diacritics inserter? Works doesn't have macrons, and it's annoying to be copy-and-pasting from the internet.

Is it not possible to type macrons with the US international keyboard layout? I admit I've never tried, but if it isn't I can add buttons for macroned characters. I'm pretty sure it's possible to type all the other accented characters if your layout is configured right.

[faiuwle]

Looking over the cluster-making functions (which I haven't messed with in a looong time), there's nothing there that would necessarily cause a crash, but they are recursive so it could just be that if you give it too much to work with the function takes too long to execute and the OS decides that the program must have encountered an infinite loop and axes it. If you get crashing from making clusters, try reducing the number of consonants in each set (e.g. instead of doing {10 consonants} + {10 consonants} +.... do {5 consonants} + {5 consonants} +.... etc., until you get all the clusters you want). I'll try to rewrite it in a non-recursive way, though.

[Qwynegold]

In the phonotactics section, it's impossible to remove allowed onsets, codas, etc. once they've been entered.

You have to use the cluster-making stuff at the bottom of the page; make the clusters/singletons that you want to remove, select onsets/peaks/codas and click "Remove".

Even if I add diphthongs to the allowed peaks, the program puts a syllable break between them when adding words. When adding words with a coda cluster, only the first consonant is displayed, even if I have defined that cluster in the phonotactics.

Are you clicking "Save Phonotactics" every time you update it? Otherwise it won't save, although the letters that are actually displayed won't change back.

Ah, both things work now.

Is it possible to have the program skip phonotactics?

Well, it no longer represents words as strings of phonemes, so it's more annoying. I could have a way to simply allow all possible phonotactics, which would result in all phonemes being analyzed as being in the onset of the first syllable, and then you would have to sort it out by hand later.

Yeah, then I would only need to add the syllable breaks myself.

So if I have it add for example all consonants + all consonants + all consonants + all consonants + all consonants + all consonants as an allowed cluster, the program crashes.

Thanks for letting me know; I'll look at it. In Windows at least, the program just hung when I did that rather than crashing, but in any case that's not what it should do.

I think the problem is that there's around 387,420,489 possible clusters, which is just too much for the computer to handle.

[faiuwle]

Yeah, then I would only need to add the syllable breaks myself.

You would actually have to add everything yourself (designations of onset/peak/coda, supras, etc.). In fact, I think it might actually be easier if no phonemes were assigned by the program at all in that case, since you would probably wind up having to delete all the existing phonemes and reenter them into the correct slots. Also, since the possibilities for onset/peak/coda are now infinite the dialog setup I have now wouldn't work either. Something like this will probably be necessary eventually though, so I might as well set it up now.

I think the problem is that there's around 387,420,489 possible clusters, which is just too much for the computer to handle.

It's possible; I don't yet have a good intuitive understanding of exactly how many shenanigans the average PC can handle at one time, or how much of a burden recursion is on RAM. I'll try reimplementing it as a loop anyway and see if it performs any better.

[dhok]

Looking over the cluster-making functions (which I haven't messed with in a looong time), there's nothing there that would necessarily cause a crash, but they are recursive so it could just be that if you give it too much to work with the function takes too long to execute and the OS decides that the program must have encountered an infinite loop and axes it. If you get crashing from making clusters, try reducing the number of consonants in each set (e.g. instead of doing {10 consonants} + {10 consonants} +.... do {5 consonants} + {5 consonants} +.... etc., until you get all the clusters you want). I'll try to rewrite it in a non-recursive way, though.

Adding {n}+{t} made it crash.

[faiuwle]

Just {n} + {t}? Or other combinations as well? Do /n/ and /t/ show up ok if you click on them in the phonology tab?

[dhok]

Not just {n}+{t}, I don't think. It crashes when I try to save the phonology, but they're fine on their own.

[faiuwle]

You mean it crashes when you click "Save Phonotactics"? Do you have the box checked so that it updates all the words too?

[dhok]

I'll check it right now.
Even if I do have the box checked...it crashes. And it isn't just /nt/.

[faiuwle]

It also crashes if you don't have the box checked?

[dhok]

Yessir. It crashes no matter what the cluster is, or whether I check the box.

[faiuwle]

Alright, I found the bug. Now I just have to reboot into the linux to fix it in the main source code and recompile everything. I don't know how that one got through...

[faiuwle]

Ok, should be fixed now, just redownload.

8/29/09
- Fixed bug where program would crash when saving phonotactics involving coda clusters

I still don't know about the cluster-making algorithm, so for now at least I'll just work on a way to override the phonotactics and just set everything manually. I'll have to write a new GUI.

[dhok]

I work Windows XP, and although it is a magnificent program, I'd like to report a bug: the program crashes when I try adding coda clusters.

Can I get more specifics here? Adding 3x3x3 coda clusters works fine on my end - are you trying to add {lots of consonants} + {lots of consonants}...x5 or something, like in Qwynegold's case?

Also, can we get a diacritics inserter? Works doesn't have macrons, and it's annoying to be copy-and-pasting from the internet.

Is it not possible to type macrons with the US international keyboard layout? I admit I've never tried, but if it isn't I can add buttons for macroned characters. I'm pretty sure it's possible to type all the other accented characters if your layout is configured right.

It would just be easier to have a menu for those...it shouldn't be too hard to make, right? All of those accents...also, what if I wanted, say, s-with-macron for /s:/...

[faiuwle]

It's not a matter of it being hard to make, it's a matter of coming up with a design that isn't horribly clunky. Trapping keystrokes to create my own input method for the program would prevent you from entering that combination of keystrokes as literals, and anyway it feels like reinventing the wheel since you can just change your layout anyway. Adding buttons where you click on the one with the character you want and the character appears in the proper box is easy to use, but there would have to be one of those buttons for every character for every place anyone would want to input characters like that. Reimplementing character map is not very attractive either. Having something where you select text and then click a button to transform relevant characters into accented characters seems like the best solution, but again, there'd have to be one of those for every input area, although I could probably get away with only one per tab. Maybe it would work as a free-floating window. It's all GUI though, so while it's easy, it won't be quick. Any suggestions you have as far as what would be easiest to use would be nice to have before I start in on it.

Accents are limited to vowels because, AFAIK, there aren't any that can apply to all Latin-1 characters without using combining diacritics, and I don't want to mess with having to parse those out.

[Qwynegold]

Yeah, then I would only need to add the syllable breaks myself.

You would actually have to add everything yourself (designations of onset/peak/coda, supras, etc.). In fact, I think it might actually be easier if no phonemes were assigned by the program at all in that case, since you would probably wind up having to delete all the existing phonemes and reenter them into the correct slots. Also, since the possibilities for onset/peak/coda are now infinite the dialog setup I have now wouldn't work either. Something like this will probably be necessary eventually though, so I might as well set it up now.

Yeah, I figured that out now. Would it be possible to have just a checkbox that says something like "All phonotactics allowed", and when it's checked, the program just automatically generates a phonetic transcription, based on what letters you have defined for different phonemes? Then you'd get to change the transcription if you want. But one needs to be allowed to type it in oneself, instead of choosing alternatives from boxes like it is now. Which of these methods of changing the transcription you'd get to use, would depend on wheter or not you have that "all phonotactics allowed" box checked.

Oh, and I noticed one more problem. When changing the transcription of a word, when you get to choose a phoneme or cluster from a box, if you have a lot of possible clusters, the box won't fit on the computer screen, and you can't scroll it.

I think the problem is that there's around 387,420,489 possible clusters, which is just too much for the computer to handle.

It's possible; I don't yet have a good intuitive understanding of exactly how many shenanigans the average PC can handle at one time, or how much of a burden recursion is on RAM. I'll try reimplementing it as a loop anyway and see if it performs any better.

Really?? In any case, it might be too much for a human to define all of those. Because, if I do for example {a e i o u} + {a e i o u} + {a e i o u}, I would get diphthongs such as aaa, aae, eei, ioo, etc, and would have to manually remove all that have two or more of the same phoneme in a row.

[faiuwle]

Yeah, I figured that out now. Would it be possible to have just a checkbox that says something like "All phonotactics allowed", and when it's checked, the program just automatically generates a phonetic transcription, based on what letters you have defined for different phonemes? Then you'd get to change the transcription if you want. But one needs to be allowed to type it in oneself, instead of choosing alternatives from boxes like it is now. Which of these methods of changing the transcription you'd get to use, would depend on wheter or not you have that "all phonotactics allowed" box checked.

The program actually encodes more information that you can actually see in the phonetic transcription, so just being able to type that in wouldn't really work. I am going to make something like that were you are restricted if you check the right box, though.

Oh, and I noticed one more problem. When changing the transcription of a word, when you get to choose a phoneme or cluster from a box, if you have a lot of possible clusters, the box won't fit on the computer screen, and you can't scroll it.

How many do you have? I can definitely fix that though - it just never occurred to me to worry about it.

[zelos]

Exacly what does the Supersegmentals and phonotactics do here?

[Torco]

suprasegmentals allows you to add, well, suprasegmental features to your phonology: gemination, for instance. then, in the word input interface, you can add gemination to what vowels you want to.

phonotactics in more interesting: one you enter a phonotactical system the program will automatically do it's best to predict the syllabification based on those rules, so if you enter a CV system and you input "helo" it will get together the word as "he.lo", while if you enter a VC system, it will probably do something like "h'el'o

[Qwynegold]

Yeah, I figured that out now. Would it be possible to have just a checkbox that says something like "All phonotactics allowed", and when it's checked, the program just automatically generates a phonetic transcription, based on what letters you have defined for different phonemes? Then you'd get to change the transcription if you want. But one needs to be allowed to type it in oneself, instead of choosing alternatives from boxes like it is now. Which of these methods of changing the transcription you'd get to use, would depend on wheter or not you have that "all phonotactics allowed" box checked.

The program actually encodes more information that you can actually see in the phonetic transcription, so just being able to type that in wouldn't really work. I am going to make something like that were you are restricted if you check the right box, though.

Oh, OK.

Oh, and I noticed one more problem. When changing the transcription of a word, when you get to choose a phoneme or cluster from a box, if you have a lot of possible clusters, the box won't fit on the computer screen, and you can't scroll it.

How many do you have? I can definitely fix that though - it just never occurred to me to worry about it.

According to my calculations, I had 196 polyphthongs when I encountered that problem.

[Torco]

bump... this just can't get pruned.

[Herra Ratatoskr]

Nuts! Nutsnutsnutsnutsnuts! I was literally just about to post that I was going to be working on a dictionary program, then I see that faiuwle has a) beaten me to the punch (and by a long shot at that!) and b) is offering many more features that I had planned for mine.

So I guess I'll ask a) faiuwle how the program is coming, and b)the rest of the ZBB if they'd be interested in me posting about my planned dictionary program. It would be somewhat different from faiuwle's, in that it would be web based, powered by PHP and so be more something people could integrate into their website, so there might be room for both programs in the conlang community. Any interest?

P.S. and sorry about any thread-jacking.

[Hakaku]

So I guess I'll ask a) faiuwle how the program is coming, and b)the rest of the ZBB if they'd be interested in me posting about my planned dictionary program. It would be somewhat different from faiuwle's, in that it would be web based, powered by PHP and so be more something people could integrate into their website, so there might be room for both programs in the conlang community. Any interest?

Nothing wrong with providing an alternative, I think it only benefits the community even more. If it's web-oriented, then I'll reassure you that people would def be interested.

[Torco]

well, faiuwle's hasn't got as man features as he promises, since he's not done developing it. a web based dictionary would be awesome. the more the merrier !