Hi Stewart,
It's good to know you are using SenseClusters, please do feel free to
let us know of any further questions, comments, etc. as they arise.
Let's see what we can do about your current queries...
On Mon, Oct 6, 2008 at 3:57 PM, Saunders, Stewart <[EMAIL PROTECTED]> wrote:
> Dear Prof. Pedersen;
>
> I am using SenseClusters1.01 and am getting good results. I do have two
> questions for which I cannot seem to find the documentation:
>
> 1) What is the command for running SVDPACKC? I have two input files,
> lap2 and Prefix.harbomat, and I want to run them on the SVDPACKC software to
> get lao2 and lav3. I have tried using svdrun and las2.c as commands, but
> they don't work, even in the c shell. I know it is possible because the
> wrapper program gives me the results. The README for svd doesn't give me
> that information.
The program that actually runs SVDPACKC is called las2, and that is
distributed with SVDPACKC and should be found in the /External
directory along with the other SVDPACKC files...
The program mat2harbo.pl is meant to be a "helper" program that gets
your data into the format required for SVDPACKC (Harwell-Boeing) and
then svdpackout.pl is another "helper" program that gets your output
from SVDPACKC into a human readable form.
So, you may want to try all of those in combination, mat2harbo.pl
followed by las2, followed by svdpackout.pl. This is more or less the
sequence that the driver program discriminate.pl runs...in fact,
here's a little excerpt from discriminate.pl that shows how these are
run....
$numform = "5$format"; ## numform is 5f16.XX
$status=system("mat2harbo.pl --numform $numform --param
$svd_string $svdin > matrix");
die "Error while running mat2harbo.pl on <$svdin>\n" unless $status==0;
system("las2");
$harbomat="$prefix.harbomat";
$status=system("mv matrix $harbomat");
die "Error in creating <$harbomat>\n" unless $status==0;
# reconstruction
$status=system("svdpackout.pl --rowonly --format $format lav2
lao2 > $svdout");
die "Error while running svdpackout.pl\n" unless $status==0;
Note that you can't run the above directly as is, but it's meant to
give a general idea of how discriminate.pl runs SVDPACKC....
You can see more about mat2harbo and svdpackout here...
http://search.cpan.org/~tpederse/Text-SenseClusters-1.01/Toolkit/svd/mat2harbo.pl
http://search.cpan.org/~tpederse/Text-SenseClusters-1.01/Toolkit/svd/svdpackout.pl
>
> 2) I assume that the order of the word vectors in the file,
> prefix.wordvec (the rows) is the same as the order of the features in the
> Prefix.features file. Is this correct?
Yes! *wordvec is a feature by feature matrix that shows the counts of
the two features (represented by the rows and columns) ocuring
together. The features that make up wordvec for shown in the *features
file.
> Also, is the order of the columns
> in Prefix.vectors(not using svd) the same as the order of the words in
> Prefix.features?
Yes! *vectors shows how each context has been represented in terms of
the features selected (so it is a context by feature representation,
where the number of columns is that found in *features).
>
> I have clustered the instances and the words(features) with good results.
> What I wish to do now is use the word vectors for factor analysis.
Great! Sounds like an interesting project - do keep us posted on how things go.
Cordially,
Ted
>
> Thanks so much for your time,
>
> Stewart Saunders
>
> Purdue
>
>
--
Ted Pedersen
http://www.d.umn.edu/~tpederse
-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
senseclusters-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/senseclusters-users
