Using %gra tier to extract transitive and intransitive relations

26 views
Skip to first unread message

Liam Considine

unread,
Feb 8, 2012, 4:16:56 PM2/8/12
to chibolts
Hello chibolts community,

I am currently trying to extract (with reasonable recall and
precision) instances of transitivity and intransitivity from the
CHILDES transcripts.

I have been creating a .cut file with a list of dependency searches. I
have a question about how to most efficiently use this method.

Here is what i believe to be some of most basic transitive relations
translated into dependency queries (please help me if i am mistaken!):

1|0|ROOT^2|1|OBJ
1|0|ROOT^2|3|DET^3|1|OBJ
1|0|ROOT^2|4|DET^3|4|MOD^4|1|OBJ
1|2|SUBJ^2|0|ROOT^3|2|OBJ
1|2|SUBJ^2|0|ROOT^3|5|DET^4|5|MOD^5|2|OBJ
1|2|SUBJ^2|0|ROOT^3|4|DET^4|2|OBJ
1|2|SUBJ^2|0|ROOT^3|4|MOD^4|2|OBJ
1|3|SUBJ^2|3|AUX^3|0|ROOT^4|3|OBJ
1|2|SUBJ^2|0|ROOT^3|5|DET^4|5|MOD^5|2|OBJ
1|3|SUBJ^2|3|AUX^3|0|ROOT^4|5|DET^5|3|OBJ
1|4|SUBJ^2|4|AUX^3|2|NEG^4|0|ROOT^5|4|OBJ

not so sure about:
1|3|OBJ^2|3|SUBJ^3|0|ROOT
as in question "what did I forget about?" response "the dog you hit."

---------------------------- here is where my question comes into play
-------------------------------------

I am interested in all instances of a verb taking an object, they do
not necessarily have to be the first things that appear in a line of
transcript.

Do i need to transpose each of these searches to different sentential/
dependency depths?
What i mean is if i start with: 1|2|SUBJ^2|0|ROOT^3|2|OBJ
Do i need to add these? 2|3|SUBJ^3|0|ROOT^4|3|OBJ & 3|4|SUBJ^4|0|
ROOT^5|4|OBJ etc....
Would these transposed searches find the dependency relationships i am
interested in places other than sentence initially?

Many thanks for all the energy and advice!

-Liam Considine



Brian MacWhinney

unread,
Feb 9, 2012, 9:47:44 PM2/9/12
to chib...@googlegroups.com, Franklin Chen, Kenji Sagae, Mark Johnson
Dear Liam,

It seems to me that the really best way to do this type of searching would be to use
the XML version of the CHILDES database and to write XQuery search strings. One reason for this
is that the identity of the ROOT as a verb is directly available in the XML for each word. Admittedly the root
is usually the verb anyway, but that would make things a bit more reliable. However, the
big issue is the one you mention in the bottom of your message.
You don't really want to be depending on the numbers to guide this
search. For example 1|0|ROOT^2|1|OBJ is the same for your purposes as 2|0|ROOT^3|2|OBJ.
So, you just want to find string where the OBJ has a pointer that matches the serial number of
the ROOT. To do this you would want variables in a search string.
I have never done this myself and we have not yet configured any of our XML search facilities to
do this, but it seems to me that this is the right approach in the long run. Franklin Chen has
been writing some stuff of this type. If you know how to program, you could tackle this or perhaps
get some help. It is certainly an interesting and very general problem. I am copying to
a few other people who may also find this interesting.

-- Brian MacWhinney

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

Reply all
Reply to author
Forward
0 new messages