ASR N-best lists are provided for the 1-side test and train condition.

NOTE: N-best lists are not broken out for the 10 and 30 second
      condition, nor are the two sides of a conversation combined for
      the 1 conversation condition.

Filename convention:

  Speech Waveform ::  <segmentName>.sph
      N-best list ::  <segmentName>.nbest

N-best lists are located in the directories:

  ./train/asr_nbest
  ./test/asr_nbest

NOTE: There are known instances of waveforms that are basically void
      of speech.  These segments will have empty or near empty 
      n-best list files.

Format of the n-best list file:
-------------------------------

File format consists of words in a hypothesis followed by the
hypothesis id, with the number after the colon indicating the index in
the n-best list.  

   Note that this file is "uncleaned" in that we still have 'compound'
   words such as 'YOU_KNOW' and some of the spellings are not
   normalized.  The hypothesis in the ctm file is cleaned with
   compounds broken up and spellings normalized.

   Example for ./nbest/xbgi.nbest


I_I THINK IT HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:1)
I I_I THINK IT HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:2)
I_I_I THINK_THAT HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:3)
I_I_I THINK I'VE HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:4)
I_I_I THINK THEY HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:5)
I_I_I THINK THAT HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:6)
I_I THINK_IF I HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:7)
I_I_I THINK I HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:8)
I_I THINK THEY HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:9)
I_I_I THINK YOU HAD_A LITTLE SHORT TERM INFLUENCE ON_IT BUT (xbgi-00001:10)