Carnegie Mellon University
Informedia Digital Video and Spoken Language Document Testbed

Corpus of MPEG encoded video (broadcast news, documentaries, government public information). Some data is fully and accurately transcribed, other with errorful automated transcripts, some with both. Segmentation pointers available, as are various levels of summaries for each segment, including one-line capsules, representative sequence of bit-map images, video skims.

1,500 hours of video data, MPEG-1 encoded, one terabyte.

Most of the content is copyrighted, available for experimentation but not available for subsequent distribution.

Further information

For further information about this testbed, see: http://www.informedia.cs.cmu.edu.

Researchers who wish to make use of this testbed should contact: Scott Stevens at Carnegie Mellon.

prepared by William Y. Arms
last modified 1/4/99 wa/cr