Page 1 of 2

Do the trees explain word order?

Posted: March 23rd, 2015, 12:47 am
by Stephen Hughes
Do the syntax trees contain information about the syntax of various sentence types and their word order?

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 1:27 am
by Stephen Carlson
Perhaps Jonathan can answer in more detail, but I've noticed that the trees group syntactically related words together into constituents without preserving the word order. Whether and/or how the original word order is somehow indicated in the trees, I don't know.

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 7:36 am
by MAubrey
No current trees, no.

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 8:07 am
by Jonathan Robie
The word order is preserved in the trees using an attribute that indicates the position of each word, the attribute is currently called morphId (I think 'position' would be a better name). You can sort on that token to restore the original order of a sentence. I do this in my query results for the plain text that occurs above a tree.

Word order is preserved in general, but not for post-positives and constructs that require changes in order in order to show the structure. When this happens, the discontinuous attribute is added to flag the change in word order. When you see that attribute, it's worth looking at the morphId.

You can see both morphId and discontinuous attributes in the following fragment:

Code: Select all

      <wg nodeId="410030100020120" class="cl" role="s">
         <w morphId="41003010002" class="conj" lemma="γάρ" discontinuous="true">γὰρ</w>
         <wg nodeId="410030100010110" class="cl">
            <w morphId="41003010001" class="adj" role="o" lemma="πολύς" case="accusative"
               gender="masculine"
               number="plural">πολλοὺς</w>
            <w morphId="41003010003" class="verb" role="v" head="true" lemma="θεραπεύω"
               person="third"
               number="singular"
               tense="aorist"
               voice="active"
               mood="indicative">ἐθεράπευσεν</w>
            <pu>,</pu>
My stylesheets don't currently show morphId or the discontinuous attribute to the user. I'm trying to simplify the presentation. I do think about showing the discontinuous attribute in the stylesheet, and might do that if I find a presentation that I like.

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 8:45 am
by Jonathan Robie
Stephen Hughes wrote:Do the syntax trees contain information about the syntax of various sentence types and their word order?
Constituents within clauses can contain labels for 's' (subject), 'v' (verb), 'vc' (verb complement), 'o', 'o2' (objects), 'io' (indirect object), 'p' (predicate), 'adv' (adverbial clause). The order of these constituents is generally preserved in the trees (and the discontinuous attribute identifies the places where it is not). Micheal Palmer and I are using this in the work we are doing this year, trying to systematically generate examples that illustrate how the Greek verb works for the purpose of language instruction.

So if I understand your question, the answer is "yes".

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 10:43 am
by Stephen Hughes
Jonathan Robie wrote:
Stephen Hughes wrote:Do the syntax trees contain information about the syntax of various sentence types and their word order?
Constituents within clauses can contain labels for 's' (subject), 'v' (verb), 'vc' (verb complement), 'o', 'o2' (objects), 'io' (indirect object), 'p' (predicate), 'adv' (adverbial clause). The order of these constituents is generally preserved in the trees (and the discontinuous attribute identifies the places where it is not). Micheal Palmer and I are using this in the work we are doing this year, trying to systematically generate examples that illustrate how the Greek verb works for the purpose of language instruction.

So if I understand your question, the answer is "yes".
Let me raise you. That infotmation is derived from the accidence and is not syntax per se. I look at these trees and see grammar relationships strangely explained in situ. The paper I gave you in the airport with the dependencies of various elements. That "takes a genitive" etc. is half of syntax. The other half is what I call the balances which I've mentioned from time to time. I don't find syntax explained in the trees.

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 1:41 pm
by Jonathan Robie
Stephen Hughes wrote:That "takes a genitive" etc. is half of syntax. The other half is what I call the balances which I've mentioned from time to time. I don't find syntax explained in the trees.
These labels go beyond "takes a genitive", they identify the role of the various parts of a clause with respect to a verb, and they involve human judgement.

Can you be clearer about what you are looking for beyond labels like 's' (subject), 'v' (verb), 'vc' (verb complement), 'o', 'o2' (objects), 'io' (indirect object), 'p' (predicate), 'adv' (adverbial clause)? What would an "explanation of syntax" look like to you?

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 4:22 pm
by MAubrey
When I hear 'explain word order,' my understanding of the question is this:

Do the tree give a motivation for why the linear ordering of Greek constituents is what it is?

The answer to that question is 'No.'

That's because we're now talking about the difference between what has been traditionally referred to as 'configurational languages' versus 'non-configurational languages' better: 'discourse configurational'). Basic English trees provide a motivation for linear constituent order because that order is directly tied to grammatical relations (and in turn, propositional structure). The order is meaningful and that meaningfulness is marked in the trees. But the motivations for Greek word order aren't like that. Greek word order is motivated by information structure: assumptions about the mental representation of participants by the speaker about the hearer (topics, foci, assertions, & predication.) in conjunction with the interface between syntax and prosody.

No current treebanks account for these concepts. It isn't even really clear if it is possible for such trees to account for this stuff.

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 4:44 pm
by Jonathan Robie
MAubrey wrote:Do the tree give a motivation for why the linear ordering of Greek constituents is what it is?

The answer to that question is 'No.'

!!! SNIP !!!

No current treebanks account for these concepts. It isn't even really clear if it is possible for such trees to account for this stuff.
Ah, now I understand, and I agree. On the other hand, treebanks might be helpful to someone doing the kind of research needed to answer these questions.

Re: Do the trees explain word order?

Posted: March 23rd, 2015, 6:04 pm
by Stephen Carlson
Stephen Hughes wrote:Let me raise you. That infotmation is derived from the accidence and is not syntax per se. I look at these trees and see grammar relationships strangely explained in situ. The paper I gave you in the airport with the dependencies of various elements. That "takes a genitive" etc. is half of syntax. The other half is what I call the balances which I've mentioned from time to time. I don't find syntax explained in the trees.
Looks like you would prefer dependency tree instead of constituency trees. Other treebanks do the former, e.g. PROIEL.