From gerkan at umich.edu  Fri May 25 09:33:38 2007
From: gerkan at umich.edu (Gunes Erkan)
Date: Fri May 25 08:55:55 2007
Subject: [Mead] Fwd: Fwd: LexRank threshold
In-Reply-To: <160ddd660705241506o53a7c23g9f0a832ab534c33@mail.gmail.com>
References: <46cedea10705231549l612d4816ia5f3e313cf61d12c@mail.gmail.com>
	<4656080a.0dd091a7.2dfb.ffffdef2SMTPIN_ADDED@mx.google.com>
	<160ddd660705241506o53a7c23g9f0a832ab534c33@mail.gmail.com>
Message-ID: <46cedea10705250633y463a78bfla6cc47829a691078@mail.gmail.com>

---------- Forwarded message ----------
From: Anthony Fader <afader@umich.edu>
Date: May 24, 2007 6:06 PM
Subject: Re: Fwd: LexRank threshold
To: radev@umich.edu
Cc: Gunes Erkan <gerkan@umich.edu>, clairlib-dev@umich.edu


Hi,

It looks like there is no way to control the similarity threshold as a
parameter, but you can easily modify the script to do this. Edit the
script in MEAD_HOME/bin/feature-scripts/lexrank/getcos.pl. Replace
these lines:

$cosine = GetLexSim ($l,$m);
print "$i $j $cosine\n";
print "$j $i $cosine\n";

with:

my $thresh = 0.1; # or whatever value you want
my $sim = GetLexSim ($l, $m);
if ($sim > $thresh) {
print "$i $j $sim\n";
print "$j $i $sim\n";
} else {
print "$i $j 0\n";
print "$j $i 0\n";
}

Let me know if this doesn't work, and I will try to find a better way.

--Tony

On 5/24/07, Dragomir R. Radev <radev@umich.edu> wrote:
> TOny or Josh?
>
> >
> > Can someone help with this question. I know -jump option is NOT the thresho=
> > ld,
> > but is there a thresholding option in MEAD?
> >
> > -gunes
> >
> > ---------- Forwarded message ----------
> > From: Kaustubh Raosaheb Patil <kap@liacc.up.pt>
> > Date: May 23, 2007 10:31 AM
> > Subject: LexRank threshold
> > To: gerkan@umich.edu
> >
> >
> > Dear Erkan,
> >
> > I am using MEAD-3.10. I want to perform some simple experiments with
> > LexRank by changing threshold value. Is the -jump option equivalent to
> > threshold (the one in the JAIR article)?
> >
> > Thank you and regards,
> > Kaustubh Patil
> >
> > --
> >
> > Knowledge can only be got in one way, the way of experience;
> > there is no other way to know.
> > - Swami Vivekananda
> >
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> >
> > Kaustubh R. Patil
> > Research Assistant,
> > LIACC,
> > Artificial Intelligence and Data Analysis Group,
> > University of Porto,
> > Rua de Ceuta 118-6=BA,4050-190, Porto,
> > Portugal
> >
> > Phone : (00351) 22 33 92 090 {ext: 31}
> > Fax   : (00351) 22 33 92 099
> > Email : kap <at> liacc.up.pt
> > Web   : www.liacc.up.pt/~kap
> >
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> >
> >
>
>
> --
> Drago
>