For example, a noun phrase might break up into these ways:

You should eat [noun phrase]
You should eat fish
You should eat the fish
You should eat the fresh fish

You should eat NP(N)
You should eat NP(Det, N)
You should eat NP(Det, Adj, N)

Is there a finite number of possibilities for the way a noun phrase might be constructed? Or is there an infinite number of possibilities?

I'm wondering because I'm starting to see papers that talk about machine learning techniques for noun-phrase detection, but if there are a finite amount of combinations, then it seems someone would have a complete list of rules or something out there.

What do you think?

  • Just in English or generally? You need to add the nlp tag anyway but please also add the english tag if that's what you're asking. Feb 8 '14 at 5:33

A noun phrase is only what you (or the linguist you learn from) define it to be. Different understandings of a language could possibly lead to different definitions of what constitute a noun phrase.

Given that you have some precise idea of what constitute a noun phrase, you then have to express that idea so that you can communicate it to others. To communicate it precisely, you need to express it with some kind of formal (mathematical ?) structure or notation.

There are many ways to do that, a simple one being phrase structure grammars introduced by Chomsky half a century ago.

Such a grammar is constituted of rules describing how phrases (a noun phrase for example) are built by putting together (sub-)phrases and parts of speech (words).

So for example you could add to your lists of examples

You should eat NP(NP, Prep, NP)

for the sentence

You should eat fish from the blue sea

This is an example of a noun phrase composed with two other noun phrases "fish" and "the blue sea", and a preposition, according to a recursive rule. Examples can be more complex and involve different kinds of phrases.

Such recursive rules can allow you to make arbitrarily long noun phrases, and thus you can have an infinite number of noun phrases, though all defined with a finite number of rules (of ways to construct them): I added only one rule to your set.

Now, whether you have an infinite or a finite number of phrases of a given type depends on the way you understand your language and wish to describe it. In general, people do understand languages as having an infinite number of noun phrases, which can be characterized by a finite number of rules.

I insist on the fact that it all depends on the linguist's understanding because there is not "a complete list of rules or something out there".

Actually there are usually many such lists for a given language, corresponding to different understandings of the structure of the language, corresponding to different linguistic theories. It does not show too much for simple noun phrases, but the complete description of a language goes far beyond that.

Furthermore, there are also different ways of expressing the structure of the language, with rules or otherwise. The world is a complicated place.


Actually, it can be easily proven that there are an infinite number of possible noun phrases, AND also that that infinite set can be generated by a finite number of (recursive) rules.

  • I think this is almost certainly correct, but you need to provide some arguments to show it, usually via links. Feb 8 '14 at 9:35
  • + @hippietrail It is well known, and often used, that an infinite set can be generated with a finite number of rules. But I doubt that you can prove so easily "that there are an infinite number of possible noun phrases" without being tautological. Because before stating anything about noun phrases, or any other kind of phrases, you have to define what they are. I am not disputing the conclusions you claim to prove, but rather the fact that you claim it to be a provable property of a concept given a priori, rather than a consequence of your own (arbitrary?) definitional choice for that concept.
    – babou
    Feb 8 '14 at 22:59
  • @babou: I'm not making any claims. I'm just trying to encourage bta to improve their answer so readers can find it satisfying. Feb 9 '14 at 2:19
  • 1
    @hippietrail Sorry. Problem with site constraints. I was really commenting the answer, but as you approved it, I thought it might be a good idea to copy you too. The "+" at the beginning is intended to mean "you too" since the author of the question gets it anyway.
    – babou
    Feb 9 '14 at 2:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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