Predictive text toy

Topics that can go away
User avatar
Xwtek
Posts: 720
Joined: Mon Sep 24, 2018 3:35 am

Re: Predictive text toy

Post by Xwtek »

Vijay wrote: Thu Nov 21, 2019 9:56 am I don't know of anyone in a Spanish-speaking country named "Andreas," though.
Well, I just type something on google translate.
IPA of my name: [xʷtɛ̀k]

Favourite morphology: Polysynthetic, Ablaut
Favourite character archetype: Shounen hero
chris_notts
Posts: 682
Joined: Tue Oct 09, 2018 5:35 pm

Re: Predictive text toy

Post by chris_notts »

bradrn wrote: Thu Nov 21, 2019 4:47 pm
Xwtek wrote: Thu Nov 21, 2019 9:25 am It's completely broken, though if I try to use Indonesian text on it.
GF-2 wrote: Halo, nama saya Andreas. Saya adalah murid kelas 5 SD Mahardika Bandung.


Ang nabarap-kapangkat wa aba hereng di nasa kopinda ang dapat tama kaya ang kandidato at penkata. Kinabuha saad walang pekeng gumawa kami. Alam rin nang, sagunang mga aking mahulang kami ng pagtosdaman ang kahit ano kakamali sa buong kalayaan. Kapag naglalahan na siyang bahagi andar lahat ng pinagtugtupay ng aking geoli ng gumagabangga. Just at mang pang alagaat di naman kaming tarek
For some reason, the output looks like it's Filipino. (Although I need Filipino natives to confirm this).
I don’t speak Filipino either, but this definitely looks like Filipino to me as well, mainly because of all the ang and ng.
Also mga (plural). Although the general feel of it seems a bit off. I wonder if actually it's mixed together a few Austonesian languages somehow? When I tried it on Spanish it randomly switched back to English half-way through.
bradrn
Posts: 6261
Joined: Fri Oct 19, 2018 1:25 am

Re: Predictive text toy

Post by bradrn »

chris_notts wrote: Fri Nov 22, 2019 1:43 am When I tried it on Spanish it randomly switched back to English half-way through.
It does tend to do that for some reason. Maybe it was mainly trained on English-language webpages?
Conlangs: Scratchpad | Texts | antilanguage
Software: See http://bradrn.com/projects.html
Other: Ergativity for Novices

(Why does phpBB not let me add >5 links here?)
Vijay
Posts: 1248
Joined: Sun Jul 08, 2018 9:13 am
Location: Austin, Texas, USA

Re: Predictive text toy

Post by Vijay »

I'm pretty sure it's not real Tagalog.
User avatar
Pabappa
Posts: 1359
Joined: Sun Jul 08, 2018 11:36 am
Location: the Impossible Forest
Contact:

Re: Predictive text toy

Post by Pabappa »

https://talktotransformer.com/

At >$2500 per month, I can no longer sustain this site without a sponsor. If you'd like to present your product or service here, please email me.

Nice. I was wondering where the money was coming from. Up till now he's just been hosting the site out of his own generosity. I hope he can find a sponsor or maybe some generous donors because even though I think I've pretty much seen what the program can do I still go back to it once a week or so to try out something new.

Also, there was a notice up earlier that the output had been shortened. This seems to have become permanent now, unless perhaps they can get more bandwidth if they receive more funding for the site.
Travis B.
Posts: 6858
Joined: Sun Jul 15, 2018 8:52 pm

Re: Predictive text toy

Post by Travis B. »

At >$2500 per month, I can no longer sustain this site without a sponsor. If you'd like to present your product or service here, please email me.

You'll have to remember that I'm not a good writer and I'm not even trying, but I've written around a dozen hours of material, and maybe most of it has been going on for six months now. If you want to host it yourself, I'd be happy to send you a work-space permit.

This site will also appear on BleedingCool.com. I need a website
Yaaludinuya siima d'at yiseka wohadetafa gaare.
Ennadinut'a gaare d'ate eetatadi siiman.
T'awraa t'awraa t'awraa t'awraa t'awraa t'awraa t'awraa.
User avatar
Pabappa
Posts: 1359
Joined: Sun Jul 08, 2018 11:36 am
Location: the Impossible Forest
Contact:

Re: Predictive text toy

Post by Pabappa »

looks like this is a paid product now, though the pricing is friendly towards end users (as low 8 cents per use) rather than large corporations who want to host instances of the tool themselves. Also, its possible that the code is being used by someone somewhere for free since it's open source.

The site owner also notes that newer stronger technology is now available, and though it is *not* open source, it has a pool of ~100 times as much text to base its output on, and thus can produce more realistic results.

At some point the modules may become so realistic that people prefer the older less efficient kind ... sort of an uncanny valley phenomenon but for perhaps a different reason.
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

With apologies where they are due.
More: show
Prompt:

Do you know how few people study Middle Egyptian? Very few. Yet I'm one of them and the most prominent one on facebook.

Completion:

How do you think it feels, you naive polytheist? I studied it in 1991 as a child when I was going to school at Al-Azhar and the Masjid Saida in Cairo. I have a decent proficiency in both Ptolemaic and Egyptian hieroglyphics. At the time I was fascinated by the culture and its presence in the hieroglyphs. When you study hieroglyphics in any historical context they become internalised and become a part of the history and society itself, as some people and cultures are still heavily dependent on it today. I had a three-dimensional paper map of Egyptian towns on my study mat, but it was mostly in the language of Ra. This was how I learned it. Ra was still Egyptian, I knew it better than any Egyptian one, but it was unclear how and where it was written or what it meant. I spent many hours at the floor of the dining hall at Al-Azhar until I became clear about its meaning.
Prompt:

Here's a conlang by me but also by someone else with the initials that sound like "victor". The part this person made is actually a "pseudo-conlang", a "re-lex" of English. I'll make some changes so that it's a conlang.


Completion:

The other ideas

"pseudo-conlang":

I don't know how to do a real conlang.

Unicode symbols I want to keep (if possible)

Should I add "random" to every possible string in lexical scope (like most conlangers do)?

I don't know how to do a real conlang.
Last edited by sasasha on Fri Jul 03, 2020 8:31 pm, edited 2 times in total.
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

For the sake of fairness, I will now rip myself.
More: show
Prompt:

It's a 'conscience' marker: a reminder of responsibility, and it's like a weight you can't shake. Vegeco superimposes ethical judgments on actions at a syntactic level: it's a thought experiment conlang tailored to a particular set of ethical principles. I can say precisely none of that in Vegeco! Thread pending.

Completion:

I have played with Vegeco for about 5-10 minutes, and I don't want to do so much more. It's too abstract. It's too murky.

And yet, if you like complex conlangers, this might be a good way to explore your own biases and biases. It's an exercise that may be difficult to take to a stack of paper, but might not be altogether impossible.

Conlang design, meanwhile, seems to be moving toward models based on free agency and this is all good. And while it can be fun to think about conlangers and to tweak their tones and features, it seems to me a useful avenue of study when thinking about conlangs, as well.

In any case, I'm trying to learn Vegeco.

Questioning the system

Even if we are forced to continue to craft new conlangs (which I hope won't ever be necessary), we may find a way to extend the foundation of existing ones. For instance, if all we have is a data base of patterns (and patterns are like moods) then we might add mechanisms that allow for pattern-defining, which, in my experience, can be crucial to the illusion of moods. This is a long shot, but I wonder.
Last edited by sasasha on Fri Jul 03, 2020 6:38 pm, edited 1 time in total.
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

The tool is a true philosopher of conlangery.

Interestingly, the next sentence was this:

"But while I am optimistic about the possibiltiy for such artificial neural networks, I am not optimistic about"

The AI started to talk about itself, essentially.

Now I must stop, because I have my debit card connected to this thing and it's horrifyingly addictive.
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

This was also very entertaining:

1. Harley Davidson – American by Birth. Rebel by Choice.
2. Volkswagen – Think Small.
3. Porsche – There is no substitute.
4. Aston Martin – Power, beauty and soul.

5. Zagato – Designing for real life.
6. BMW – House of great people, place where cars and life go hand in hand.
7. Mercedes – Enjoying the ride.
8. Audi – One of the world's most admired brands.
9. Lotus – Taking on the world.
10. Lamborghini – Across the world – From Tokyo to Madrid.
11. BMW i3 – With its radical design.
12. Ferrari – From racing legends to the future.
13. Ferrari – Driving play.
14. Caterham – Downtime with flair.
15. Bentley – Your adventure.
16. Jaguar – A legend in every sense of the word.
17. Koenigsegg – Fast. Over the moon.
18. Lamborghini – Style, Speed, Vandalism.
User avatar
alice
Posts: 962
Joined: Mon Jul 09, 2018 11:15 am
Location: 'twixt Survival and Guilt

Re: Predictive text toy

Post by alice »

Have you tried giving it a phoneme inventory and seeing what morphology and syntax it gives back?
Self-referential signatures are for people too boring to come up with more interesting alternatives.
User avatar
Raphael
Posts: 4568
Joined: Sun Jul 22, 2018 6:36 am

Re: Predictive text toy

Post by Raphael »

sasasha wrote: Fri Jul 03, 2020 2:01 pm 18. Lamborghini – Style, Speed, Vandalism.
Wouldn't that be more appropriate for the first entry in the list?
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

alice wrote: Fri Jul 03, 2020 2:06 pm Have you tried giving it a phoneme inventory and seeing what morphology and syntax it gives back?
You owe me $0.08 ;)
More: show
sasasha and InferKit wrote:
Phonology
Consonants
/tʷ/ /tˤ/ /tʲ/ /kʷ/ /kˤ/ /kʲ/ /ʔ/
/tʷʰ/ /tˤʰ/ /tʲʰ/ /kʷʰ/ /kˤh/ /kʲʰ/
/ʃʷ/ /ʃˤ/ /ʃʲ/ /ʍ/ /ʕ/ /ç/
/tʃʷ/ /tʃˤ/ /tʃʲ/
/tʃʷʰ/ /tʃˤʰ/ /tʃʲʰ/
/m/ /mˤ/ /n/ /nˤ/ /ŋ/ /ŋˤ/
/l/ /lˤ/ /ɾ/ /ʁ/
/w/ /j/


Vowels
/ɨ/ /ɨː/ /ʉ/ /ʉː/
/ɛ/ /ɛː/ /œ/ /œː/ /ɔ/ /ɔː/
/ɐ/ /ɐː/


Orthography
Consonants
du dd di gu gg gi
tu tt ti ku kk ki
su ss si hu hh hi
ju jj ji
cu cc ci
m mm n nn q qq
l ll r rr
w y


Vowels
é éé ó óó
e ee ø øø o oo
a aa


(C)V(l,r,w,y,m,n,q)
Pharyngealised onsets pharyngealise codas which have a pharyngealised variant.


Morphology



pharyngealised

Pharyngealised variants are declined exactly as if they were monophthongs. A terminal marker is added in front of the vowel or syllable stem if the underlying consonant is [x] or [v] (or, if it is not, of the initial consonant if it is).

Click here to listen to a recording of Azz-lîhātî_logu (an example)


References


American Heritage Dictionary of the English Language

Compiled by W. L. Rudman and R. N. Whitehead

New York: The Macmillan Company

1978

Copperfield, Catherine (1994) 'Phonology' in Webster's Third New International Dictionary of the English Language, B. Webster (ed.)

Garden City, NY: Doubleday Anchor Books

1246, ISBN 0-393-00979-4

Collins, C. D. (1995) 'Phonology' in Webster's Third New International Dictionary of the English Language, B. Webster (ed.)

Garden City, NY: Doubleday Anchor Books

1246, ISBN 0-393-04395-2
It cheated a bit. Either that or this language is very light on syntax...
Last edited by sasasha on Fri Jul 03, 2020 6:36 pm, edited 1 time in total.
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

I ran it again, changing the end of the input to "Morphology & Syntax" and discovered a bunch of interesting stuff about consonantal diphthongs.
More: show
There are four cases for any particular morphology. Each of them is also the name of a doublet, for example, all Pharyngealised syllables are diphthongs. A diphthong can be divided into two parts, the 'dominant' consonant and the 'derived' consonant. The first part is that with the open syllable, that is represented by [d]. The second part is the 'derived' consonant, that is represented by the following triples:

*r in English (e.g. murmur /mʊɪr/)

*ɟ in the phonological family (e.g. [ɟ] in Italian),

*h in the orthological family (e.g. [h] in Russian),

*m in English (e.g. [m] in Italian)


Each of the triples has a pattern of 'long', 'close', 'open' or 'proximal' penetration. The meaning of the morphophonemic regularisation (as described in Phonological / Grammaticalisation ) is as follows:

*long: penetrates the nasal line and midline of the preceding syllable, with the consequent changes in phonemes. *open: penetrates nasal, midline and rostral line. *proximal: penetrates r, l, m.

These are actual cases:

*Pharyngealised diphthong: pp ɪ p ɢ pp *Diphthong with closure: p s t ɪ s t p *Diphthong with extended closing: p t p t p *Diphthong with open nasal, midline and rostral
(Also learned to do 'more' boxes. Woot.)
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

[plug]

If anyone likes playing with this sort of thing, by the way, perhaps you'd be interested in having a look at or contributing to my 'Semi-Automatic Poetry' blog. I should update it more, but life gets in the way!

https://bionically.blogspot.com

[/plug]
User avatar
alice
Posts: 962
Joined: Mon Jul 09, 2018 11:15 am
Location: 'twixt Survival and Guilt

Re: Predictive text toy

Post by alice »

sasasha wrote: Fri Jul 03, 2020 6:34 pm I ran it again, changing the end of the input to "Morphology & Syntax" and discovered a bunch of interesting stuff about consonantal diphthongs.
"Dominant" consonants "penetrating r, l, m"? I think you may be onto something here.
Self-referential signatures are for people too boring to come up with more interesting alternatives.
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

alice wrote: Sat Jul 04, 2020 9:44 am
sasasha wrote: Fri Jul 03, 2020 6:34 pm I ran it again, changing the end of the input to "Morphology & Syntax" and discovered a bunch of interesting stuff about consonantal diphthongs.
"Dominant" consonants "penetrating r, l, m"? I think you may be onto something here.
I was curious how Inferkit would handle double entendre, so I ran your response through it. Sadly, it confused linguistics for music theory, and missed BDSM entirely.

More: show
"Dominant" consonants "penetrating r, l, m"? I think you may be onto something here. In other words, the constant B-flat going for the back is strongly playing against the AS, as do the times when the "C-flat" goes for the backs or rims.


This is also true of "Meta-D" ("Elemental"), but I personally don't think of that consonant as "winning."

Well, maybe not entirely... I love the 'topic soup' this one produced.
User avatar
alice
Posts: 962
Joined: Mon Jul 09, 2018 11:15 am
Location: 'twixt Survival and Guilt

Re: Predictive text toy

Post by alice »

You could probbaly work that into the musical versions of philosophical languages somehow.
Self-referential signatures are for people too boring to come up with more interesting alternatives.
sasasha
Posts: 468
Joined: Mon Aug 06, 2018 11:41 am

Re: Predictive text toy

Post by sasasha »

alice wrote: Wed Jul 08, 2020 9:37 am You could probbaly work that into the musical versions of philosophical languages somehow.
Ohmygod I wish I was still studying composition. Generating a description of a piece via this tool and then writing it to fit would get me marks, even (/especially) if it sounded crap.
Post Reply