finding dative constructions

[Rowland, Caroline]

Hi
I have done this for a couple of children but as you say it is very time consuming. I decided to search for dative verbs (compiling a hopefully near complete list from Pinker, Levin and all the others papers I could find) and then going through by hand and removing non-dative uses.  I did try searching via "to" and "for" but that was even worse! 

Caro

-----Original Message-----
From: chibolts at googlegroups.com [mailto:chibolts at googlegroups.com] On Behalf Of Brian MacWhinney
Sent: 13 June 2010 21:54
To: chibolts at googlegroups.com
Cc: Kenji Sagae; Shuly Wintner
Subject: Re: finding dative constructions

Dear Bobbi,

     I started trying your Plan B method of going through the files line by line.  Have you every tried doing this?  I think terrible things start to happen to your eyes and brain.  Seriously, I think that should really be a last resort, unless your intention is to just take a small sample of the English database.  
   Regarding the methods you are using, we know that coding of OBJ2 by GRASP is still very unreliable.  In order to improve it, we will eventually have to derive semantic features on verbs from some system such as FrameNet and try to build that into the grammatical relations tagger. For now, it is using the most superficial of cues, I am afraid.  
    I know that some other people have done work on this.  Amy Perfors has some recent stuff and Jess Gropen (1989) studied this in some detail.  
    It seems to me that this is a wheel that people are continually re-inventing and it would be really nice to provide a more solid basis for this.  I think that means identifying in some detail the verbs that are likely to take the two dative forms.  With such a list in hand, I would then add additional tags to the MOR and GRASP taggers.  I think this would then improve the OBJ2 tagging.  For the prepositional dative, a slightly different approach might be needed, but still the verb feature tagging would be central.  
   In regard to datives that are basically benefactives, marked with "for", I would have thought that retrieval of those would not be too tricky, particularly if they follow after the OBJ.  

I am copying this to a couple of other people, in case they have further ideas.

--Brian MacWhinney

On Jun 13, 2010, at 6:39 PM, RCorrigan wrote:

> I am trying to search for dative constructions in English, both
> ditransitives and prepositional object constructions, produced by all
> speakers (both adults and children). I have been using COMBO with the
> following "s" file:
> *OBJ^*^*OBJ2*
> for^*
> prep|to^*
> 
> I then hand check to weed out all the "for" and "to" constructions
> that are not datives. It also seems that for some files there are alot
> of errors with coding OBJ2, so the combo search misses alot of
> constructions.  To pick up some of the rest, I have been using KWAL
> with a list of verbs that are commonly used in dative constructions.
> Of course, I miss any uncommonly used verbs. Plus, many verbs that are
> used commonly in datives are also commonly used in other constructions
> (e.g., get and make) , so again there is a lot of hand screening. And
> in addition, some of the constructions I've already identified with
> the COMBO search show up again with the KWAL search.
> 
> All in all, I think I'm doing this in a very clunky way and it might
> actually be faster to read each line of each file by hand!
> 
> I checked the archives and I can't seem to find any posts on these
> constructions. Does anyone have any suggestions?
> 
> Bobbi Corrigan
> 
> -- 
> You received this message because you are subscribed to the Google Groups "chibolts" group.
> To post to this group, send email to chibolts at googlegroups.com.
> To unsubscribe from this group, send email to chibolts+unsubscribe at googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/chibolts?hl=en.
> 
> 

-- 
You received this message because you are subscribed to the Google Groups "chibolts" group.
To post to this group, send email to chibolts at googlegroups.com.
To unsubscribe from this group, send email to chibolts+unsubscribe at googlegroups.com.
For more options, visit this group at http://groups.google.com/group/chibolts?hl=en.

-- 
You received this message because you are subscribed to the Google Groups "chibolts" group.
To post to this group, send email to chibolts at googlegroups.com.
To unsubscribe from this group, send email to chibolts+unsubscribe at googlegroups.com.
For more options, visit this group at http://groups.google.com/group/chibolts?hl=en.