How do natural languages prevent word ambiguity in "compound words"?

Question

So for context, I am occasionally working on a sort of conlang, and asked this question just recently: How to create words which will be unambiguously parsable in a conlang? In there I run into the problem where I have navofodoroda mean either navo-fodo-roda or na-vofodo-roda, as we defined words/meanings for all those "parts". The problem is, when you write it without all that extra whitespace/metadata (of the dashes), you have navofodoroda and you can't tell which one you are using. So that got me to that question, how do you resolve the problem where I have many words in a lexicon, and then it turns out that I run into this sort of conflict way down the road...

That leads me to this question here. How do natural languages like English or Spanish (or any other) deal with creating words so they don't run into this ambiguity problem?

For example, in English we have the word nineveh but I know that that first part isn't nine for some reason. But even so, this word is unique as far as I can tell (I can't think of many counterexamples), so it's not like nineveh has two distinct meanings. I know English has some words with multiple distinct meanings ("tear", for example), but you can tell based on context what it is. But in what other ways do languages work their magic so they don't run into these sort of conflicts? Do they somehow limit the possibility/scope of possible words, so as to prevent these conflicts? If so, how do they do this? (It seems to me like you'd have to study/know every single word in the lexicon to do it appropriately, which makes it seem extremely hard and brittle). Please show how some natural language or a few natural languages deal with this problem.

The only way I can see solving this problem is to have -- like programming -- special "keywords", which are reserved and you can't use any of the related sound patterns of those words. Then that way you can prevent this ambiguity. But that seems nebulous and hard to pinpoint down to me, I don't even know if it's possible. So in my case, na- is the special prefix, so basically no words other than not can contain at all the affix na. But that seems highly limiting, that would mean na is only used like once in the language? That doesn't seem to be realistic / make sense. I don't know tho, how other languages handle this.

Answer 1

Natural languages don't do this, ambiguities happen. An example is the German word Staubecken that can be analysed in two ways: Stau-becken "pond, basin, dam reservoir" and Staub-ecken, plural of Staubecke "dust corner".

When the ambiguity becomes too annoying (i.e., the two reading can occur in the same context) speakers will chose another word to express themselves with less ambiguity.

Answer 2

How do natural languages like English or Spanish (or any other) deal with creating words so they don't run into this ambiguity problem?

As jk said, they don't. Suppose I tell you that a door is "unlockable". Does that mean it's (un-lock)-able, able to be unlocked, or un-(lock-able), not able to be locked? Sometimes it's also ambiguous which morphemes are involved. Is "entrance" from enter+ance, or from en+trance?

Natural language has a lot of redundancy in it. Look at how much you can compress English text, for example, using something as simple as Huffman encoding—English text is written in a very inefficient way that has significant redundancy built in. Ths is why yu cn rd txt wth all the intrnl vwls rmvd. Or why you can hold a conversation in whispers, where voicing information is lost. Or over a noisy channel like radio. This redundancy is a feature, not a bug, and in this case it also means that it's usually pretty clear whether "entrance" means a way to enter, or to put someone into a trance. I'd wager you've never actually run into difficulties telling which meaning of "entrance" was intended.

Now, some languages are more explicit about how noun compounds go together. Many Bantu languages, for example, have gender agreement on the connectives that link nouns to other nouns, and this agreement indicates how they connect:

ma-koki m-a yo m-a mw-ana w-a li-boso
6-right 6-of you 6-of 1-child 1-of 5-front
"Your rights as the firstborn child"

The agreement markers indicate that wa liboso "first" (literally "of the front") attaches to mwana "child", while ma mwana wa liboso "of the firstborn" and ma yo "your" (literally "of you") both attach to makoki "rights".

But these aren't strictly necessary, and many other Bantu languages and dialects get by just fine without them.