If we are to extend our Automated Tourist Guide so as to understand sentences like the one at the beginning of the chapter, and to generate appropriate responses, we need first of all to know about the formal structure of the English language. We need, that is, to be able specify all the possible forms of acceptable sentences in the English language (or at least, in the subset of English which we expect our system to handle), and to do so in a manner that enables us to write computer programs which, given strings of words as input, will distinguish between those that we feel as native speakers to be grammatical sentences in English and those that are not.
In many grammars, the syntax and semantics are strongly interlinked insofar as the meaning of a complete sentence is in part determined (according to the so-called `rule-by-rule hypothesis') by the manner in which its constituent words and phrases are syntactically related. We discuss this further in section 5.3. For the present, we wish merely to stress the fact that every well-formed sentence in a natural language has a formal syntactic structure, and that we can unambiguously describe that structure irrespective of whatever meanings it may be used to convey.
There are two strong motivations for this: in the first place, such a program would capture our intuitions as speakers of English about the formal structure of English sentences. We will want our program to be able to distinguish, for example, between novel but perfectly grammatical strings such as those in (1) below and nonsentences of the kind exemplified in (2) below:
1. a. All stuffed grey elephants are moderately inflammable.b. There are no such things as triangular virtues.
2. a. Inflammable all grey moderately elephants stuffed are.
b. There are are are are as as virtues.