I’m a middle-aged computer scientist/philosopher, who specialized in artificial intelligence and machine learning back in the stone age when I was getting my degrees. Since then I’ve done a bit of work in probabilistic simulations and biologically inspired methods of problem solving, mostly for industry. I’ve recently finished writing a book about politics, although God knows if I’ll ever sell a copy. Now I’m into a bit of everything. Politics. Economics.
I came here looking for input into a conlang project that I’m working on. Basically it involves the old Sapir-Whorf/Eprime/Loglan dream of creating a language that’s better suited for rational cognition than English, and I’m looking for linguistic mechanisms that might aid in this and that need to be built in from the bottom up (since surface mechanisms can be added later). I already know of the three conlangs mentioned above, although I don’t speak them, so I’m looking for ideas that aren’t contained therein, or that if they are might have been missed by a person without a deep knowledge of the languages. I did a search of the archives here and saw some discussion around this general topic, but nothing of immediate use, although I could easily have missed something.
I do have a conlang draft. A few thoughts based on my conlang thinking:
Loglan/Lojban is a language were math was an afterthought. That’s likely mistake.
If you look at a concept like grandfather, using the word “grand” doesn’t make much sense. I think it’s better to say something like father-one for grandfather, father-two for great-grandfather.
The same way the boss of your boss should be boss-one.
Having a grammer in which relationships can be expressed well is very valuable.
I think that loglan attempt to build on existing roots of the widely spoken languages is flawed because it allows less freedom organizing the language effectively. It would be good to have a lot of concepts with 3 letters instead of 5.
In my language draft I started to take concept of graph theory for naming relationships (the structure of the words matters but the actual word is provisional):
bei node in same graph cai node parent doi node children
beiq relative caiq parent doiq son/daughter
bei person employed in the same company caiß boss (person with authority to order) doiß direct (person who can be ordered)
Once you understand that structure and learn the new word “fuiq” for sibling, you can guess that
a direct coworker is called fuiß. Of the in a graph notes that share the same parent note are “fui”.
I like grouping concepts this way where I can go from parent to son/daughter simply by going one forward in the alphabet and replace “c” with “d” and “a” with “o” (“i” get’s skipped because the word ends in “i”).
I did use a similar principle for naming numbers:
ba 0 ce 1 di 2 fo 3 gu 4 ha 5 je 6
For the number I also gave adding a “q” meaning. It turn the number into base 16.
Base 16 numbers are later quite useful if you want to make an expression like north-east.
At the moment pilots use phrases based on the clock to navigate: “There’s a bird at 2 o’clock.” It’s much better to bake numbers more centrally into the language.
In case you haven’t seen it http://selpahi.de/ToaqAlphaPrimer.html is a nice draft for a new language. I like how the language makes every sentence end in an evidential.
In it I think he makes a mistake that he doesn’t use capital letters but non-asci character instead.
I think that it’s great that his language doesn’t follow the Lojban place system but uses prepositions like a normal language.
It sounds like you were trying to construct an a-priori conlang, in which the meaning of any word could be determined from its spelling, because the spelling is sufficient to give the word exact coordinates on a concept graph of some sort. I thought about this approach some time ago, but was never able to find a non-arbitrary concept graph to use, or a system of word formation that didn’t create overly long or unpronounceable words.
I was originally thinking about including non-ascii characters, but eventually compromised on retaining English capitals instead. The biggest problem that any conlang faces is getting people to use it, and anything that makes that more difficult, such as requiring changes to the standard American keyboard, needs to be avoided unless it’s absolutely necessary.
Hi,
I’m a middle-aged computer scientist/philosopher, who specialized in artificial intelligence and machine learning back in the stone age when I was getting my degrees. Since then I’ve done a bit of work in probabilistic simulations and biologically inspired methods of problem solving, mostly for industry. I’ve recently finished writing a book about politics, although God knows if I’ll ever sell a copy. Now I’m into a bit of everything. Politics. Economics.
I came here looking for input into a conlang project that I’m working on. Basically it involves the old Sapir-Whorf/Eprime/Loglan dream of creating a language that’s better suited for rational cognition than English, and I’m looking for linguistic mechanisms that might aid in this and that need to be built in from the bottom up (since surface mechanisms can be added later). I already know of the three conlangs mentioned above, although I don’t speak them, so I’m looking for ideas that aren’t contained therein, or that if they are might have been missed by a person without a deep knowledge of the languages. I did a search of the archives here and saw some discussion around this general topic, but nothing of immediate use, although I could easily have missed something.
All ideas welcome.
I do have a conlang draft. A few thoughts based on my conlang thinking:
Loglan/Lojban is a language were math was an afterthought. That’s likely mistake. If you look at a concept like grandfather, using the word “grand” doesn’t make much sense. I think it’s better to say something like father-one for grandfather, father-two for great-grandfather. The same way the boss of your boss should be boss-one. Having a grammer in which relationships can be expressed well is very valuable.
I think that loglan attempt to build on existing roots of the widely spoken languages is flawed because it allows less freedom organizing the language effectively. It would be good to have a lot of concepts with 3 letters instead of 5.
In my language draft I started to take concept of graph theory for naming relationships (the structure of the words matters but the actual word is provisional):
bei node in same graph
cai node parent
doi node children
beiq relative
caiq parent
doiq son/daughter
bei person employed in the same company
caiß boss (person with authority to order)
doiß direct (person who can be ordered)
Once you understand that structure and learn the new word “fuiq” for sibling, you can guess that a direct coworker is called fuiß. Of the in a graph notes that share the same parent note are “fui”.
I like grouping concepts this way where I can go from parent to son/daughter simply by going one forward in the alphabet and replace “c” with “d” and “a” with “o” (“i” get’s skipped because the word ends in “i”).
I did use a similar principle for naming numbers: ba 0
ce 1
di 2
fo 3
gu 4
ha 5
je 6
For the number I also gave adding a “q” meaning. It turn the number into base 16. Base 16 numbers are later quite useful if you want to make an expression like north-east. At the moment pilots use phrases based on the clock to navigate: “There’s a bird at 2 o’clock.” It’s much better to bake numbers more centrally into the language.
In case you haven’t seen it http://selpahi.de/ToaqAlphaPrimer.html is a nice draft for a new language. I like how the language makes every sentence end in an evidential. In it I think he makes a mistake that he doesn’t use capital letters but non-asci character instead.
I think that it’s great that his language doesn’t follow the Lojban place system but uses prepositions like a normal language.
Also, the topic is now up and running in the regular “discussion” area.
It sounds like you were trying to construct an a-priori conlang, in which the meaning of any word could be determined from its spelling, because the spelling is sufficient to give the word exact coordinates on a concept graph of some sort. I thought about this approach some time ago, but was never able to find a non-arbitrary concept graph to use, or a system of word formation that didn’t create overly long or unpronounceable words.
I was originally thinking about including non-ascii characters, but eventually compromised on retaining English capitals instead. The biggest problem that any conlang faces is getting people to use it, and anything that makes that more difficult, such as requiring changes to the standard American keyboard, needs to be avoided unless it’s absolutely necessary.