The file trecvid2007.dev.asr_v002.tar.gz contains the final version of output
from the University of Twente's automatic speech recognition engine applied
to the 2007 Sound and Vision (search and feature task) development data.

Note for the final version for the development set:

- a few files for the development set are empty - this may be due to
   lack of speech and is being investigated.

- one xml file for each video in our own xml format

- one xml file for each video in mpeg7 format

- a tar file with all lattices (in case teams want to do language model rescoring)

- a short text file explaining the three file formats

- Here is a citation to use:

Marijn Huijbregts, Roeland Ordelman and Franciska de Jong, Annotation
of Heterogeneous Multimedia Content Using Automatic Speech
Recognition.  in Proceedings of SAMT, December 5-7 2007,
Genova, Italy

in Bibtex format:

@inproceedings{ TRECVID:ASR,
  author     = {Marijn Huijbregts and Roeland Ordelman and Franciska de Jong},
  title      = {Annotation of Heterogeneous Multimedia Content Using Automatic Speech Recognition},
  booktitle  = {Proceedings of the second international conference on Semantics And digital Media Technologies ({SAMT})},
  series     = {Lecture Notes in Computer Science},
  month      = {December},
  address    = {Berlin},
  publisher  = {Springer Verlag},
  year       = {2007}
}



The file trecvid2007.test.asr_v002.tar.gz contains the final ASR for the
2007 Sound and Vision (search and feature task) test data.

If you have questions about the ASR you can contact:

ordelman@ewi.utwente.nl
marijn.huijbregts@gmail.com