4.2.2 The Syntax Rules we will use

The implemented syntax rules.

Luckily, there's a simple trick that will make a limited form of coordination available to us. We'll add an auxiliary set of categories named np2, np1, v2, v1, etc. These auxiliary categories allow us to specify left-recursive rules to a certain depth of recursion. For example, the rules which have something to say about NPs will be replaced by the following:

A DCG allowing only limited recursion.

np2--> np1.
np2--> np1, coord, np1.
np1--> det, noun2.
np1--> pn.

Similarly, the rules controlling nouns will become:

                          
noun2--> noun1.
noun2--> noun1, coord, noun1.
noun1--> noun.
noun1--> noun, pp.
noun1--> noun, rc.

While this is a rather blunt way of dealing with the problem of left recursion in a grammar, it enables us to parse the examples we want without having to implement a more sophisticated parser.

Another shortcoming of these rules should be mentioned. As you might have noticed by now, we've imposed limits on inflectional morphology---all our examples are relentlessly third-person present-tense. This is a shame, since tense and its interaction with temporal reference is a particularly rich source of semantic examples. Nonetheless, we shall not be short of interesting things to do.

englishGrammar.pl: View Download

But for all its shortcomings, this small set of rules (to be found in englishGrammar.pl) assigns tree structures to an interesting range of English sentences:

``Mary loves every owner of a siamese cat.''

``John or Mary smokes.''

``Every man that loves a woman visits a therapist.''

``John does not love a therapist or woman that smokes.''

``If a therapist talks then a man works.''

If you want to test the grammar, you will have to do this within a semantic framework, e.g. -calculus: lambda([mary,knows,every,owner,of,a,siamese,cat],Sem).