Re: [Mead] Sentence weighting

From: radev@umich.edu
Date: Thu Sep 28 2006 - 08:34:50 EDT


Your question is a bit ambiguous.

Here is how mead works:

1. it extracts features using the feature extractors
2. it combines them using the combiner. This is where you can specify
how the features should be combined, e.g., linearly, loglinearly, etc.

By default, MEAS uses a linear combination of features using default
weights. Note that the Length is not included by default. It is a
special "cutoff" value that removes sentences that are too short.

Drago

>
> --===============1354292284==
> Content-Type: multipart/alternative; boundary="0-1546600298-1159441513=:83743"
> Content-Transfer-Encoding: 7bit
>
> --0-1546600298-1159441513=:83743
> Content-Type: text/plain; charset=iso-8859-1
> Content-Transfer-Encoding: quoted-printable
>
> Dear all,
> =20
> I am newbie in summarization. I am trying to device a coherent text summ=
> arizer. For comparision I selected MEAD (version 3.07), though it doesn't=
> claim to produce coherent extracts. One thing I noted in MEAD that the s=
> entence weights are linear combination of, 1. distance from centre 2. pos=
> ition 3. length.=20
> =20
> So can anyone please tell me how the position and length based weighing =
> works?
> =20
> Any suggestions appreciated.
> =20
> stay abstract,
> Kaustubh
> =20
> =09
> ---------------------------------
> Get your email and more, right on the new Yahoo.com=20
> --0-1546600298-1159441513=:83743
> Content-Type: text/html; charset=iso-8859-1
> Content-Transfer-Encoding: quoted-printable
>
> Dear all,<br> <br> I am newbie in summarization. I am trying to device a =
> coherent text summarizer. For comparision I selected MEAD (version 3.07),=
> though it doesn't claim to produce coherent extracts. One thing I noted =
> in MEAD that the sentence weights are linear combination of, 1. distance =
> from centre 2. position 3. length. <br> <br> So can anyone please tell me=
> how the position and length based weighing works?<br> <br> Any suggestio=
> ns appreciated.<br> <br> stay abstract,<br> Kaustubh<br> <p>&#32;
> <hr size=3D1>Get your email and more, right on the <a href=3D"http://us=
> .rd.yahoo.com/evt=3D42973/*http://www.yahoo.com/preview"> new Yahoo.com</=
> a>=20
>
> --0-1546600298-1159441513=:83743--
>
> --===============1354292284==
> Content-Type: text/plain; charset="us-ascii"
> MIME-Version: 1.0
> Content-Transfer-Encoding: 7bit
> Content-Disposition: inline
>
> _______________________________________________
> Mead mailing list
> Mead@lists.si.umich.edu
> http://lists.si.umich.edu/mailman/listinfo/mead
>
> --===============1354292284==--
>
>

-- 
Dragomir R. Radev                    Associate Professor
SI, CSE, Ling                     U. Michigan, Ann Arbor 
http://www.eecs.umich.edu/~radev         radev@umich.edu              
_______________________________________________
Mead mailing list
Mead@lists.si.umich.edu
http://lists.si.umich.edu/mailman/listinfo/mead



This archive was generated by hypermail 2b30 : Tue Jun 09 2009 - 05:00:08 EDT