[Corpora-List] Subcat Questions
Chris Brew
cbrew at ling.ohio-state.edu
Wed Jun 18 16:59:25 UTC 2003
On Wed, Jun 18, 2003 at 03:19:06PM +0100, peetm wrote:
>
> Sorry, if this would have been better asked elsewhere - but I thought I'd
> give it a go here.
>
> Subcategorisation Frames - can anyone point me a 'The Dummy's Guide to' - or
> provide an explanation of them, which contains only plain and simple English
> please?
>
OK I'll try. If this is too simple, sorry. If too complex,
ask. Subcategorization frames are a component of a system of ideas
designed to describe verbs in such a way as to bring out interesting
commonalities. They are based on notions such as
transitive/intransitive that you may have been taught at school, but
go further.
1) Some verbs must (or nearly must) be intransitive
John faints
we might say the subcategorization frame for this is
np _
where np means noun phrase. The np is the only thing
needed to fill in all the crucial roles of the event
of fainting. In predicate calculus one might write
faints(john)
2) Other verbs must (or nearly must) be transitive
John likes pizza
(The old men) like (the dark green trees)
the brackets mark the phrases. Linguists
have decided for independent reasons that the
phrases "the old men" "the dark green trees" can
function in the same way as the individual words
"John" and "pizza". If this is news to you, you
probably want to read the first chapter of a good
introduction to linguistics.
the frame for both "likes" sentences would be
np _ np
If you say just "John likes" it feels as if something
has been left out. So there are two obligatory things (complements)
that
needed to be added to complete "likes". In predicate
calculus one might write
likes(john, pizza)
3) Some verbs can be either transitive or intransitive
John moves
np _
John moves the books
np _ pp
3) Some verbs are yet more complicated
John gives Mary (the books)
np _ np np
John gives (the books) (to Mary)
np _ np pp
(here pp is prepositional phrase, another kind of phrase,
since "to" is a preposition)
John promises Mary (the books)
np _ np np
John promises Mary (that he will return the books)
np _ np s
(Here s stands for sentence. Quite why "that he will
return the books" should be a sentence is beyond the
current scope, but this is broadly accepted)
4) There don't seem to be frames with more than four
elements, and the only common one with that many is
John bet Mary (half a crown) (that he would return the books)
np _ np np s
5) Arguments and adjuncts.
For purposes of counting, we don't include things that
are inessential to the meaning of the verb, so
John likes pizza (on Tuesdays)
is given the frame
np _ np
and nobody worries about the "on Tuesdays". The reason is (loosely)
that "on Tuesdays" could be added to almost any sentence, and so
shouldn't count as part of the description of the verb. Things that
are part of the frame get called "arguments" and things that are not
get called "adjuncts". There is much debate about this distinction,
from discussion of whether a particular thing is better analysed as
argument or adjunct, through studies of whether machines (or indeed
humans) can reliably distinguish arguments from adjuncts to
philosophical diatribes on whether such a distinction is meaningful in
the first place. If this part strikes you as flaky, you're in very
good company. To really understand the motivations behind this you
probably need a whole course in linguistics.
More information about the Corpora
mailing list