3

I'm trying to write a program that generates valid English sentences, but without specifying what structure the sentence takes.

I want that to be determined by starting at a simple sentence S (NP VP? ) and then mutating it by using grammar rules into a new sentence (NP VP ->NP VP NP) .

Is there a finite set of rules I can use? I've searched the way and I cant find a definitive set of rules. Everything I've found seems to caveat it with 'is only a simple set of rules'. Do we only have a simple set to work with? Is there a definitive set of rules?

I believe the rules I'm talking about are called Categorial Rules...Hear are some I found:

a. S → NP Modal VP
b. VP → V AP PP
c. AP → ADVP A
d. ADVP → ADV
e. PP → P NP
f. NP → D N

Does an exhaustive list of these exist?

  • 1
    You can get a summary (with examples but without details) at Haj Ross's English syntax rules. And a partially overlapping set of alternation rules and the verb classes that license them in Beth Levin's English Verb Classes and Alternations (verb class list here, details and bibliography in the book). – jlawler Feb 22 '14 at 18:04
  • The title of this question is misleading. The set of phrase structure rules in a grammar is finite by definition. – Ceasar Bautista Feb 22 '14 at 19:02
  • @CeasarBautista not being an expert...I didn't know. – Pureferret Feb 22 '14 at 19:22
  • It's okay, but you should change it to be more clear to help people with similar questions in the future. – Ceasar Bautista Feb 22 '14 at 19:23
  • @CeasarBautista but what if they have the same misconception? I'll leave it for the comments/Answer to correct me. – Pureferret Feb 22 '14 at 19:25
4

In order to have a definitive list of the syntax rules for English, you first need to know in a definitive way what English is.

One problem is probably that you do not agree with your next door neighbor on the English-ness of all sentences you may want to check. I am not even considering vocabulary which varies greatly from person to person.

If you take a diachronic view of the language (that means analyzing the history of the language) you notice that syntactic structures change over time. But there is no big boss saying at some point: today we decide that adverbs will always be used in such and such way, with all English speakers suddenly following the new rule. For one thing there was no radio, TV or Internet to organize that.

So the evolution was progressive, the language evolving as new people were taking the place of older ones. But at no time is there a precise characterization of the structure of the language. No one knows exactly was is dead, what is still used, what is coming into being.

I would think that that alone will garantee that no single list can pretend to be a definitive description of the language.

This being said, you can always use some set of rule that will be a good approximation of the language as you perceive it. In general, choosing the rules is already the beginning a theory for the language structure.

But there is only so much information about the language you can put in the rules. The phrase structure rules you are using are, so called, "context free". That mean they do not offer much facility for carrying information such as agreement between parts of the sentence. It can be done, up to a point, but at the expense of increasing considerably the number of rules. Fundamentally, you need other mechanisms, or other techniques than context-free rules, if only to keep to reasonable size the number of rules of your grammar.

The rule are thus only a skeleton for the syntax of the language. Such skeleton is always composed of a finite set of rules. However, as I said above, it can be completed by other mechanisms (attibute, features,...) that may often (not always) be seen as if you extended the set of rules, possibly in an infinite way. But that is always only a view of it, because the specifications one uses are necessarily always expressed finitely.

Then there are languages with an internal structure of sentences that may not even be compatible with this type of syntax specification. See for example the question: What is an example of a syntactic structure that can't be represented by a BNF grammar?

And I am skipping semantics ("the bone ate the dog").

You will have your hands full, even with "only a simple set of rules".

But I would suggest you look at an introductory book on computational linguistics before you start your project.

  • I'm sure that my next answer will be about the 'skeleton' grammar rules you mentioned – Pureferret Feb 23 '14 at 0:21
1

In Syntactic Structures (1957), Chapter 5, Chomsky begins by explaining that he doesn't know whether a Phrase Structure Grammar is capable of describing English, but rejects it as an adequate theory because there are instances where the explanations are extremely complex. I'm not sure if one has been developed since then, but I suspect the answer to your question is no in virtue of the limits of PSGs.

  • "the answer to your question is no"...No they aren't finite, or No a PSG for the english language does not exist? – Pureferret Feb 22 '14 at 19:28
  • 1
    My belief is that no adequate PSG for English exists. You might find something that is good enough for your task, but I suspect nobody has developed one that could reasonably be said to describe English. – Ceasar Bautista Feb 22 '14 at 19:36
  • 2
    Oh, yes. One has been developed since then. Several times. There are now many versions, from Generative Semantics through Case Grammar, Space Grammar, Construction Grammar, Frame Grammar, Relational Grammar, Standard Theory, Arc Pair Grammar, Revised Standard Theory, Revised Extended Standard Theory, Barriers, Government and Binding, and Minimalism, to Bare Phrase Structure. To name only a few of the sects. Ever since Chomsky proved it was possible, the way to make a name in linguistics has been to make up your own theory with its own name. – jlawler Feb 22 '14 at 22:21
  • 1
    @jlawler Make that an answer. That's very informative! – Ceasar Bautista Feb 22 '14 at 22:42

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.