4

Right now I am working on this, a list of words to translate into the conlang. At first I started by manually deriving the words from Hebrew/Arabic/Sanskrit/Greek/Latin/English/Spanish, but then found it to be too tedious and time-consuming, so I started just randomly assigning many of the words to a set of precomputed words which fit a specific pattern. The patterns are (c = consonant, v = vowel):

  • cv (2 letters, like "do")
  • cvc (3 letters, like "bol")
  • ccvc (4 letters, like "gran")
  • cvcc (4 letters, like "bord")
  • ccvcc (5 letters, like "start")
  • cvcvc (5 letters, like "tekal")

Everything except the 2-letter words is a "base" word form, which starts and ends with a consonant. Then there are these rules:

  • suffix -a: make it into a noun
  • suffix -i: make it into a verb
  • suffix -o: make it into a "feature" (modifier/adjective)
  • suffix -aha: make noun into plural
  • suffix -iha: make verb into continuous (like "-ing")
  • suffix -iho: make verb into past tense (like "-ed")
  • suffix -ihi: make verb into future tense (like "will x")
  • suffix -oho: make feature into an adverb (like "-ly")

Then you can start combining words to make more complex words. There are things like -tion (capturing the result of action, written as xom by itself), or -er (the person performing the action, written as pum by itself). So you have for example:

  • meki: create
  • mekixoma: creation
  • mekipuma: creator

This seems all fine and well, with little ambiguity so far. But then I start running into trouble with the following I think.

  • prefix na-: meaning not.
  • nav: number 9 (happens to start with na-, but not prefixed with na-).
  • fod: "fold"
  • rod: "path" or "road"

So you can have the word phrase "ninefold path" as <navo> (feature/adjective) <fodo> (feature/adjective) <roda> (noun), or navofodoroda.

All is fine and well so far. We separate the words with their POS tag basically.

But now say we add a word like vofod (cvcvc), which isn't derived from "fold" (fod), but means voice. And say we want to say "without voice path", that would be the same then! navofodoroda. So now there is ambiguity!

Darn, how do I avoid this? I already have a pretty strict word-formation structure/system, but it seems to break down even then. I don't get how to avoid such ambiguity.

Perhaps I could put constraints on what letters can be used to create words, but that seems like a nightmare, how do I know what to prevent and where? In that case, no word other than na could start with or contain na-. But that seems quite restrictive, I don't know. Do other natural languages do such a thing (or conlangs)?

If I do that, how many different ways am I going to have to put constraints on word structure, and given what? What are the key words that will restrict other word formations? I have no idea if this is the right track to go down... Any other suggestions?

It is like fourier analysis, it has the word four in it, but it doesn't mean "four" the number. Somehow you are able to tell what it should mean. Or "dominican republic", you have the name "dom" (dominic), but don't think of this as containing it. And the word 'republic' which contains "re" and "public", but it has nothing to do with re + public directly. What are some other examples like this? I can't think of many? Or "nineveh" (assyrian empire), does not mean "nine" + "veh". But we don't have a same word "nine+veh" that means something else, so we never encounter this problem I'm solving. Hoping for some guidance on how to solve this.

4
  • 1
    Is there a reason you can't put spaces between words? (so "navo fodo roda" and "na vofodo roda" would be distinct)
    – Cecilia
    Dec 21, 2021 at 19:02
  • @Richard I guess that is one solution (doh), but I was going for like the greek thing where they prefix a- to mean not, or english un- or dis-. How can I make it work without separating the words so I can get those greek-derived-like English words?
    – Lance
    Dec 21, 2021 at 19:27
  • 1
    conlang.fandom.com/wiki/… Dec 22, 2021 at 17:00
  • I figured it out! A special reserved word to act as separator (toward the end of this section of text).
    – Lance
    Dec 24, 2021 at 1:19

2 Answers 2

3

Many languages use prosody to indicate separation between words. In English, for example, almost all words have a single primary stress in them. So when you hear another primary stress, you know there's been a word boundary; this is how you can tell the difference between "greenhouse" and "green house". In Latin, similarly, every word has a single ictus, which appears at a well-defined position following certain rules. Given a stream of syllables in Latin, the ictus will tell you unambiguously where to divide them into words.

Many languages also have explicit marking when nouns are combined into larger noun phrases. This might be a case marking, like in Ancient Greek and Sumerian, or it might be an extra word (often called a "particle"), like in Japanese and Swahili, or it might be a phonological change, like in Akkadian and Egyptian.

Finally, natural languages always have a lot of redundancy in them (which is a feature, not a bug). Most arbitrary strings of phonemes are not meaningful, and in general, changing the grouping of phonemes in a sentence will not produce a meaningful result. (For example, "meaningful" has meaning, "me ningf ull" does not.)

1
  • 1
    Side note, Sumerian's not the best example for case marking because it attaches to a phrase rather than a word. But I wanted to highlight different ways of doing it.
    – Draconis
    Dec 24, 2021 at 22:41
3

It may worth to have a closer look at Lojban. It has a lot of devices implemented to ensure that any legal stream of sounds can be uniquely parsed into words and that compound words can be uniquely split into components. This is done by ascribing special phonotactics to each part of speech, and by adding devices like spoken quotation marks.

Lojban does not prevent semantic ambiguity, however: it allows the speakers to be deliberately vague or ambiguous in what they say.

2
  • Interesting! Can you describe briefly what these features look like with a few examples, as relevant to this question (like how Lojban would solve it)? I am going to look through the wiki but it seems vast and I don't know if I will encounter what you are describing without a full studying of its features.
    – Lance
    Dec 22, 2021 at 21:46
  • When you can read German, the German Wikipedia article on Lojban gives you a good glimpse on Lojban: de.wikipedia.org/wiki/Lojban
    – Sir Cornflakes
    Dec 23, 2021 at 0:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.