Minutes from September 13, 2005 Teleconference

Attendees:
    NIST: Jon F.
    LDC: Stephanie S., Meghan G., Kevin W.
    ICSI: Adam J.
    KU: John M.
    ELDA: Khalid C.
    CMU: Susi B.

Topics discussed focused on data:

- KU/CHIL data reported by John M.:

A brand new data set is to be collected for the 2nd CHIL Evaluation Run. This
database will be called “Seminar’05”. It will consist of a set of audiovisual
recordings of seminars (like in the “Seminar’04” database) and of 
highly-interactive small working groups seminars.  The data will be used
for both RT-06S and internal CHIL evaluations.

The recordings will be done accordingly to the “CHIL Room Setup” specification.

The data will be collected by the CHIL partners who have set up a CHIL room at
their facilities:

 * UKA-ISL will collect at least 20 seminars before M16 (April 2005);
 * ITC-IRST will collect 5 interactive seminars before M16;
 * UPC will collect 5 interactive seminars before M20 (August 2005);
 * RESIT will collect 5 interactive seminars before M20;
 * IBM will collect 5 seminars before M20;
 * CMU will collect 5 interactive seminars before M20.

Discussion:  There was a distinction between seminars and interactive seminars.  
The consensus was that we need to listen to some data to know if the interactive
seminars will count as conference room seminars from '06.  The discussion led
to a action item for ELDA to release a sample if interactive seminars on the web.

Dev test set will be constructed by ELDA and NIST will help with data selection.

Re-release of RT-05S lecture room data:  KU/NIST/ELDA will conspire to
re-release the RT-05S.  The release will include beamformed data from
the reference speaker position tags.

- CMU data reported by Susi B.

CMU/ISL would contribute two meetings each with 4 speakers:
   close talk microphones (countryman)
   2 table microphones
   4-8 ceiling mounted video cameras.  Calibrated with the checker
       board.  mpeg-2.

- AMI data (commented by John M.)

From discussions with Thomas H. that AMI has collected data but he was unsure
about the status of the data.   Jon F. will contact them about the status.

- NIST data (in email)

NIST will donate two conference style meetings.

- ELDA comments:

* Djamal Mostefa will be handling the transcription/annotation of CHIL data.
* They will release a sample of interactive seminars.


Other items:

Adam J.:  How do they get bug fixes sent back to the producers of the data?
Khalid said ELDA was working on bug submission but there isn't a system in
place yet.



Action items:
1: Sample of interactive seminars released by ELDA
2: Release the RT-05S lecture room evaluation data
3: Jon F. will contact VT to see if they will provide data
4: Jon F. will contact AMI to see if they will provide data