#!/bin/bash
  
  # This script operates on a data directory, such as in data/train/,
  # and writes new segments to stdout. The file 'segments' maps from
  # utterance to time offsets into a recording, with the format:
  #   <utterance-id> <recording-id> <segment-begin> <segment-end>
  # This script assumes utterance and recording ids are the same (i.e., that
  # wav.scp is indexed by utterance), and uses durations from 'utt2dur', 
  # created if necessary by get_utt2dur.sh.
  
  . ./path.sh
  
  if [ $# != 1 ]; then
    echo "Usage: $0 [options] <datadir>"
    echo "e.g.:"
    echo " $0 data/train > data/train/segments"
    exit 1
  fi
  
  data=$1
  
  if [ ! -s $data/utt2dur ]; then
    utils/data/get_utt2dur.sh $data 1>&2 || exit 1;
  fi
  
  # <utt-id> <utt-id> 0 <utt-dur>
  awk '{ print $1, $1, 0, $2 }' $data/utt2dur
  
  exit 0