# Readme:  June 2013
# TrecVID 2013 automatic transcriptions
# copyright Vocapia Research / LIMSI-CNRS

Of the 7221 wav files provided, the audio partitioner does not detect
any speech in 279 files, and the speech recognizer does not detect any
speech in an additional 590 files.

 nospeech.lst       partitioner  (279 files)
 nospeech_asr.lst   asr          (590 files)

Since the language of the audio data is unknown, this was
automatically identified using the Vocapia Research/LIMSI language
identification system v4.2. It is assumed that the audio file contains
speech in only one language. 

There are a total of 7429 asr xml files, however some of these may not
have any recognized speech (590 files)
This systems identifies the language and gives a language confidence
score (lconf) in the xml files.

If the LID score was 0.8 or higher, the audio file was transcribed
with the detected language. (3036 files have a lconf >= 0.8).  Most of
these files are English (2639) with only 397 in other languages.

 list: lid_08+.lst

The remaining files were transcribed with the detected language. If
the detected language was not English (the case for 490 files of the
1465 files) they were processed twice (with the detected language and
with the English system (lang=eng lconf=1.0). For these files two
transcripts are provided, the first has the extension .xml and the
second has the extension .eng.xml.

 list: lid_08-.lst

There are 2239 files with an identified language that we do not have
an STT system for (systems are available for 14 languages: Arabic,
Mandarin Chinese, Dutch, English, Finnish, French, German, Greek,
Italian, Polish, Portugeuse, Romanian, Russian, Spanish) so these were
transcribed with the English model, but in general few words are found
and the overall confidence score is low. The files have the extension
.eng.xml. This list of files is in:

  nomodel.lst

and some are also in the list of files with no recognized words
(nospeech_asr.lst)