Crouton.org
Here's a quick demo of how to use Crouton.
This is a preliminary version of one of my V2 scripts called
V2Sum. Click on a blue line to see commentary, and click on the
comment to make it go away.
Every script consists of some import statements followed by an action.
The action in this script is wrapped in a let block so you can make
definitions, but it's still just an action. When you run a script,
Crouton assumes it contains an action and that's the work that
does.
And the action is to summarize the manuscript file. Here,
ARGS is a list of strings passed
in on the command line. So if you go to a command line and run the command
crouton V2Sum.ctn /home/marge/manuscript.psd
Crouton will set
ARGS = {"/home/marge/manuscript.psd"}
before evaluating your script. The
First function from the Prelude module
returns the first item in a list. Since there's only one action in
this sequence, we could have left out the do, but I find the script
easier to read with the do.
in
in
Here's how to use that script. First, here are the module, script,
and data files. Download all of these and put them in a folder.
Note: The file example.psd is the first few sentences from a
PPCME manuscript. example.psd ends in ".psd" so some computers think
it's a Photoshop document rather than a text file. Just save it and
load it in a text editor if you want to see what's in it.
To run the script go to a command line and run
crouton V2Sum.ctn example.psd
and a moment later, you should find a file called example-sum.psd:
((ID CMAELR3,26.4) IP-IMP)
((ID CMAELR3,26.5) IP-IMP)
((ID CMAELR3,26.6) IP-MAT (N (N apostel)) (VBP sei+t) (SV (NP-SBJ)) (SV (, QTP E_S)))
((ID CMAELR3,26.7) IP-MAT-SPE)
((ID CMAELR3,26.8) IP-MAT (PRO (PRO hit)) (BEP is) (SV (NP-LFD , NP-SBJ-RSP)) (SV (NP-OB1 E_S)))
((ID CMAELR3,26.9) IP-MAT (N (NPR Crist)) (VBP sei+t) (SV (CONJ PP NP-SBJ)) (SV (PP , QTP E_S)))
((ID CMAELR3,26.10) IP-IMP-SPE)
((ID CMAELR3,26.11) CP-QUE)
You can see that several sentences were imperative (IP-IMP), direct
speech (IP-MAT-SPE, IP-IMP-SPE) and questions (CP-QUE) that didn't fit
the first pattern in
Summary, so it
fell through to the second pattern. Other interesting things that
show up when you run this script on the whole file: The word
that can be a pronoun, a complementizer, or a determiner, but
when
that is used as a subject, it's still coded as a
determiner. There are lots of questions, imperatives, and speech.
Now, this is pretty complicated. Let me give you a quick explanation
of why I prefer this to Corpus Search 1. First, CS1 allows you to
express things like subject before or after verb pretty easily, but
since I need to split everything up depending on whether the subject
is a pronoun and what the pre-posed constituent is, etc, I found
myself writing zillions of query files to split the corpus into
different branches and ... well, the amount of programming got out of
hand in a hurry. I wanted it to be repeatable so I could give some
simple instructions if a linguist wanted to verify my results, and
with all the Makefiles (yes, I tried using
make) and the fact that
CS1 can't do "or" (now available in CS2) ... well, it got complicated.
At least as complicated as V2Sum.ctn, probably more so. And I wasn't
very confident that I was doing things right. In fact, I know I was
getting things wrong because of the
that problem. So, I
decided to try writing my own search program. I do a lot of work in
Mathematica and
Haskell and I decided to borrow some
ideas from both of them (mostly Haskell) to design Crouton.
![[CroutonLogo]](CroutonLogo1.png)
Back to main page.