Return-Path: Received: from postmark.nist.gov ([unix socket]) by postmark.nist.gov (Cyrus v2.1.11) with LMTP; Sun, 09 Nov 2003 09:30:20 -0500 X-Sieve: CMU Sieve 2.2 Received: from nist-i.nist.gov (nist1.nist.gov [129.6.94.1]) by postmark.nist.gov (8.12.5/8.12.5) with SMTP id hA9ETt18029416 for ; Sun, 9 Nov 2003 09:29:55 -0500 (EST) Received: from hawksbill.nist.gov ([129.6.13.21]) by nist-i.nist.gov; Sun, 09 Nov 2003 09:29:22 -0500 (EST) Received: from e1.ny.us.ibm.com (e1.ny.us.ibm.com [32.97.182.101]) by hawksbill.nist.gov (8.12.9-20030924/8.12.9) with ESMTP id hA9ETDau004596 for ; Sun, 9 Nov 2003 09:29:13 -0500 (EST) Received: from northrelay02.pok.ibm.com (northrelay02.pok.ibm.com [9.56.224.150]) by e1.ny.us.ibm.com (8.12.10/NS PXFA) with ESMTP id hA9ET6f8480146; Sun, 9 Nov 2003 09:29:06 -0500 Received: from d01ml604.pok.ibm.com (d01av02.pok.ibm.com [9.56.224.216]) by northrelay02.pok.ibm.com (8.12.9/NCO/VER6.6) with ESMTP id hA9ET4W1015658; Sun, 9 Nov 2003 09:29:04 -0500 In-Reply-To: <3FA8056F.7090707@nist.gov> To: Paul Over Cc: Alan Smeaton , Wessel Kraaij MIME-Version: 1.0 Subject: Re: Fw: TRECVID: reminder... notebook papers and registration X-Mailer: Lotus Notes Release 6.0.2CF1 June 9, 2003 From: John R Smith Message-ID: Date: Sun, 9 Nov 2003 09:29:05 -0500 X-MIMETrack: Serialize by Router on D01ML604/01/M/IBM(Release 6.5|September 18, 2003) at 11/09/2003 09:29:05, Serialize complete at 11/09/2003 09:29:05 Content-Type: multipart/mixed; boundary="----=_NextPartTM-000-500f68c6-dfb8-4176-8727-64f86eeaf2cf" X-MailScanner: This is a multipart message in MIME format. ------=_NextPartTM-000-500f68c6-dfb8-4176-8727-64f86eeaf2cf Content-Type: multipart/alternative; boundary="=_alternative 004FA00F85256DD9_=" --=_alternative 004FA00F85256DD9_= Content-Type: text/plain; charset="US-ASCII" Hi Paul, Please see summary as follows: The IBM team participated in all four TRECVID tasks in 2003. For shot boundary detection, the IBM results were based on the IBM CueVideo system. For high-level feature detection, we submitted ten runs based on a set of uni-modal and multi-modal runs using statistical modeling techniques such as support vector machines, multi-layer perceptrons, ensemble fusion, validity weighted model vectors, multi-nets, discriminant model fusion, and ontology-based fusion. We also created several "best-of" runs by automatically selecting best classifiers from above techniques. For story segmentation, we used an unsupervised audio-visual clustering technique that incorporated a decision-tree classifier on speech transcript. For search, we examined spoken-document retrieval and content-based techniques. The spoken-document retrieval techniques included automatic and phonetic-based approaches. We also examined spoken-document retrieval fusion-techniques over multiple matching functions and methods for re-ranking based on color blobs. We also explored techniques for fully-automatic multiple example content-based retrieval. Lastly, we explored search techniques that fuse content-based and spokent-document retrieval results based on linear weighting. Thanks, John --------------------------------------------------- Manager, Pervasive Media Management IBM T. J. Watson Research Center 30 Saw Mill River Road Hawthorne, NY 10532 (914) 784-7320; jrsmith@watson.ibm.com --------------------------------------------------- Paul Over 11/04/2003 03:00 PM To John R Smith/Watson/IBM@IBMUS cc Alan Smeaton , Wessel Kraaij Subject Re: Fw: TRECVID: reminder... notebook papers and registration John R Smith wrote: > > Hi Paul, Cedric, > > I just wanted to inform you that the IBM team will not finish our > notebook paper in time. Therefore, we are planning to bring and > distribute any writeups at the workshop instead. > > Thanks, > > John > Hi John, Alan, Wessel, and I were expecting to extract some words from the papers about the main approaches various groups took to the tasks they undertook - for use in the overviews. Since we will not be getting a paper from the IBM team, can we get something vary brief about the approaches IBM took to each task - this week?. Thanks, Paul -- Paul Over - Retrieval Group Information Access Division Information Technology Laboratory National Institute of Standards and Technology Bldg. 225 Rm. A211 (Mailstop 8940) Gaithersburg, MD 20899-8940 USA Voice: 301 975-6784 Fax: 301 975-5287 --=_alternative 004FA00F85256DD9_= Content-Type: text/html; charset="US-ASCII"
Hi Paul,

Please see summary as follows:

The IBM team participated in all four TRECVID tasks in 2003.  For shot boundary detection, the IBM results were based on the IBM CueVideo system.  For high-level feature detection, we submitted ten runs based on a set of uni-modal and multi-modal runs using statistical modeling techniques such as support vector machines, multi-layer perceptrons, ensemble fusion, validity weighted model vectors, multi-nets, discriminant model fusion, and ontology-based fusion.  We also created several "best-of" runs by automatically selecting best classifiers from above techniques.  For story segmentation, we used an unsupervised audio-visual clustering technique that incorporated a decision-tree classifier on speech transcript.  For search, we examined spoken-document retrieval and content-based techniques.  The spoken-document retrieval techniques included automatic and phonetic-based approaches.  We also examined spoken-document retrieval fusion-techniques over multiple matching functions and methods for re-ranking based on color blobs.  We also explored techniques for fully-automatic multiple example content-based retrieval.  Lastly, we explored search techniques that fuse content-based and spokent-document retrieval results based on linear weighting.

Thanks,

John

---------------------------------------------------

Manager, Pervasive Media Management
IBM T. J. Watson Research Center
30 Saw Mill River Road
Hawthorne, NY 10532
(914) 784-7320; jrsmith@watson.ibm.com

---------------------------------------------------


Paul Over <over@nist.gov>

11/04/2003 03:00 PM

To
John R Smith/Watson/IBM@IBMUS
cc
Alan Smeaton <asmeaton@computing.dcu.ie>, Wessel Kraaij <Kraaij@tpd.tno.nl>
Subject
Re: Fw: TRECVID: reminder... notebook papers and registration





John R Smith wrote:
>
> Hi Paul, Cedric,
>
> I just wanted to inform you that the IBM team will not finish our
> notebook paper in time.  Therefore, we are planning to bring and
> distribute any writeups at the workshop instead.
>
> Thanks,
>
> John
>

Hi John,

Alan, Wessel, and I were expecting to extract some words from
the papers about the main approaches various groups took to
the tasks they undertook - for use in the overviews. Since we
will not be getting a paper from the IBM team, can we get
something vary brief about the approaches IBM took to each
task - this week?.

Thanks,
Paul

--
Paul Over - Retrieval Group
                    Information Access Division
                    Information Technology Laboratory
                    National Institute of Standards and Technology
                    Bldg. 225  Rm. A211  (Mailstop 8940)
                    Gaithersburg, MD  20899-8940   USA
                    Voice: 301 975-6784    Fax: 301 975-5287                  
--=_alternative 004FA00F85256DD9_=-- ------=_NextPartTM-000-500f68c6-dfb8-4176-8727-64f86eeaf2cf--