Page 2 of 4

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 17th, 2014, 11:24 pm
by Stephen Carlson
Jonathan Robie wrote:
Stephen Carlson wrote:Can βρέχειν actually govern a genitive object? Shouldn't you use a verb's selectional restrictions (i.e. which case it governs) to identify objects?
TimNelson wrote:Stephen: My assumption (which may be wrong) is that Jonathan doesn't have that information in computer-readable format, and he's trying to see if he can get the computer to figure it out for him, rather than him having to go through and mark every verb.
Yes, precisely. I have the entire GNT in trees, and I can use that to tell me, for each verb, the cases in which it takes its objects, with a listing of each argument by case. That's information that the freely available lexicons don't give me. And if I come up with it in a useful form, I'll put it in github so it's freely available.
Well... you're going to need to be able to parse object constituents to find the head and then look at the case of the head . Since apparently you had that information, I would just use that. Recreating that information algorithmically is not really trivial and seems error-prone.

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 18th, 2014, 7:05 pm
by TimNelson
Stephen Carlson wrote:Well... you're going to need to be able to parse object constituents to find the head and then look at the case of the head . Since apparently you had that information, I would just use that. Recreating that information algorithmically is not really trivial and seems error-prone.
Hmm. But if he can recreate it algorithmically, then the algorithm may be applicable to other texts :).

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 18th, 2014, 7:36 pm
by Stephen Carlson
TimNelson wrote:
Stephen Carlson wrote:Well... you're going to need to be able to parse object constituents to find the head and then look at the case of the head . Since apparently you had that information, I would just use that. Recreating that information algorithmically is not really trivial and seems error-prone.
Hmm. But if he can recreate it algorithmically, then the algorithm may be applicable to other texts :).
I'm not so sure, unless those other texts already had someone identify the objects but not the heads of the object constituents. However, the task of identifying the objects involves identifying their heads.

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 18th, 2014, 10:19 pm
by Jonathan Robie
Stephen Carlson wrote:I'm not so sure, unless those other texts already had someone identify the objects but not the heads of the object constituents. However, the task of identifying the objects involves identifying their heads.
And I'd kind of like to be treating the nouns as the heads of noun phrases ... but there may be really good reasons to stick with the GBI scheme, especially since we are downstream from them.

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 18th, 2014, 11:06 pm
by TimNelson
Hmm. I guess my philosophy on noun phrases is that the noun is the ... maybe "head" isn't the right word [*], but the element linked to the verb, with the article specifying a mandatory attribute (definite/indefinite/qualitative), and the pronoun specifying an optional (and therefore less closely linked) attribute. Is that where you're coming from?

Going from there, I'd ask whether there isn't some way you could define a set of transforms that would turn the GBI tree into one that's the way you like it?

[*] I've realised that "head" can be used in two different ways. The first comes at things from a "tree" perspective, and the "head" is the node that links to things further up the tree (ie. verbs, in the case of direct objects). The second comes at things from a more textual perspective, and the head element is the one that has to be at the start of the phrase (ie. even in Greek, IIUC, the definite article must come before the noun it modifies). I don't know whether this is what's causing the disagreement amongst linguists; maybe someone (Mike or Stephen) might have some thoughts.

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 18th, 2014, 11:46 pm
by Stephen Hughes
Can you use the information in the dictionary to tell your algorithm which case the verb uses? There is no need to let the text speak for itself, because there is already a lot of recorded knowledge of the language that is external to the text.

When we read we use that part of our knowledge of Greek to help us work out what goes together in sentences. Is there a requirement to work from only the text itself and the work of the GBI to make these syntax-trees? You could add value to the analysis by using information from other reference works' general knowledge of the language.

Does Thayer's lexicon contain that information?

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 19th, 2014, 12:00 am
by Stephen Carlson
TimNelson wrote:[*] I've realised that "head" can be used in two different ways. The first comes at things from a "tree" perspective, and the "head" is the node that links to things further up the tree (ie. verbs, in the case of direct objects). The second comes at things from a more textual perspective, and the head element is the one that has to be at the start of the phrase (ie. even in Greek, IIUC, the definite article must come before the noun it modifies). I don't know whether this is what's causing the disagreement amongst linguists; maybe someone (Mike or Stephen) might have some thoughts.
I've never seen "head" used in your second sense, so I'd say it's not causing the disagreement between the NP and the DP hypotheses.

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 19th, 2014, 12:24 am
by Stephen Hughes
Jonathan Robie wrote:
αὕτη δὲ τοῖς δάκρυσιν ἔβρεξέν μου τοὺς πόδας καὶ ταῖς θριξὶν αὐτῆς ἐξέμαξεν
I am basically looking for a reliable algorithm to identify the case of a verb's object. If a phrase is known to be the object of a verb, what is the algorithm for identifying the case? Or should I simply rely on the GBI's head marker?
Another point about your trees is that the understood object of the verb ἐξέμαξεν is not going to show up by using this analysis, is it? Written out as it would be understood, the sentence would include the final object too, wouldn't it? ἔβρεξέν μου τοὺς πόδας καὶ ταῖς θριξὶν αὐτῆς ἐξέμαξεν μου τοὺς πόδας*

And likewise for
Luke 7:38 with the implied object spelt out wrote:καὶ κατεφίλει τοὺς πόδας αὐτοῦ, καὶ ἤλειφεν τοὺς πόδας αὐτοῦ τῷ μύρῳ.
While in this case John 11:2 καὶ ἐκμάξασα τοὺς πόδας αὐτοῦ ταῖς θριξὶν αὐτῆς would show up in an automated search, I think that not listing the understood objects (syntactically implied objects) leads to a systematic deficiency in any verb list that would be created by this method. The overall syntax of the phrase would perhaps need to intelligently looked at (perhaps sentence-by-sentence) to identify implied objects to get a more complete picture of verb usage.

People could of course claim that if it is not written in the text then it is not in the text, but clearly μου τοὺς πόδας is understood as the object of ἐκμάσσειν in the reading of the text, and in the mind of the writer, and should, in my opinion at least, be listed in a verb catalogue. That of course somewhat complicates the analysis of the data you are getting from the GBI.

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 19th, 2014, 1:53 am
by Stephen Hughes
That last post reads with a terseness the that i didnt intend for it to have. :oops: :?

Perhaps instead of those questions that I wrote hoping to raise doubts not make statements, could be ignored and other questions like "Am I right in assuming that...?" could be substituted.

Re: Genitive Pronouns in Noun Phrases in Syntax Trees

Posted: November 19th, 2014, 7:19 am
by Jonathan Robie
Stephen Hughes wrote:Can you use the information in the dictionary to tell your algorithm which case the verb uses?
If I had a machine readable dictionary that gave me this information for the cases of each verb, in a complete form, I would not need to do the query. Except, perhaps, as just a way to demonstrate the power of the query system. Which is one of the reasons for the query ...