Print

Print


>> Chapter 3 in http://lingbuzz.auf.net/lingbuzz/002821 compares different parse methods.
>  Figure 14 is a CKY parse for the structure that is nonlinear in other 
examples.

> Figure 14 shows abaS, not aS.

It's a linear structure, that's the point. I can do aS, aaS, aaaS, abaS etc., if you want any particular one. With a CKY parse they all give you a linear structure such as the one in figure 14. 

You said:

> The aS grammar can generate only lists, not more 
complex trees.

but chapter 3 explains how this only depends on the choicen parse method. 

I'm posting you here two different aS parses. I'm also changing the word-order to top-down/verb-first (to match the DOT format). First let's take a CKY parse for the sentence. Please go to http://ironcreek.net/phpsyntaxtree/ and insert this (it's best to uncheck auto subscript):

[S [A want-event] [A agent1] [A i-person] [A want-event] [A patient1] [A read-event] [A agent2] [A anyone-person] [A read-event] [A patient2] [A book] [A possessor] [A i-person] [A want-event] [A polarity] [A negation] [A read-event] [A manner] [A carelessness]]

and press draw. As you can see, I also made it explicitly an all-noun expression.

For the graph parse it has been customary (i.e. done for paaS) to compose a graph from several subphrases. This is not logically required for aS, but it keeps the graph a lot cleaner. So here's the corresponding expression for DOT - please insert it to http://graphs.grevian.org/graph and press generate:

graph{
want_event--agent1--i_person;
want_event--patient1--read_event--patient2--book--i_person;
read_event--agent2--anyone_person;
want_event--polarity--negation;
read_event--manner--carelessness;
}

So you see what we have here in this message are just word lists of different length. But it's just as difficult to see how come the graph is actually just a list, as it is difficult to see how come the list is supposed to be more than just a list. Thank goodness for computers!

What's more, you see here that semantic nouns are sufficient to express everything. Note that I had to change hyphens to underscores due to format limitations, so 'want_event' should read 'want-event'. 

In the graph we're looking at an event (head) with an agent, patient, polarity, etc. and it works just perfectly. That shouldn't be a huge surprise considering that the word verb is actually a noun. But preposition is also a noun, adverb is a noun, pronoun is a noun etc., so it's actually pretty daft to say you can't express them with nouns... as you do when you say it.

Now someone is going to say this one example does not prove all complex events can be expressed as mere lists of nouns; in fact all my lifetime would not be enough to prove it. The thing is I've now done my part and demonstrated my point. The burden of proof is on the opponent: if anyone gives one example of a meaning that can't be expressed as a list of nouns (= in aS), they should bring this example forward so we can see it :)