[egs] minor fixes to multi_en recipe (#2013)

egs/multi_en/s5/local/prepare_dict.sh

@@ -14,6 +14,15 @@ replace_cmudict_symbols=( "ah" "ah l" "ah n" )
 . ./cmd.sh
 . ./path.sh
 
+#check existing directories
+if [ $# -lt 1 ] || [ $# -gt 2 ]; then
+  echo "Usage: prepare_dict.sh /path/to/SWBD [/path/to/TEDLIUM_r2]"
+  exit 1; 
+fi 
+
+SWBD_DIR=$1
+TEDLIUM_DIR=$2
+
 # This function filters lines that are common in both files
 function filter_common {
     awk 'NR==FNR{arr[$0]++;next} arr[$0] {print}' $1 $2
@@ -38,12 +47,11 @@ for i in "${replace_swbd_symbols[@]}"; do
 done
 
 # Prepare switchboard lexicon
-local/swbd1_data_download.sh /export/corpora3/LDC/LDC97S62
-
+local/swbd1_data_download.sh $SWBD_DIR
 local/swbd1_prepare_dict.sh
 
 # Prepare cmudict + tedlium lexicon
-local/cmu_tedlium_prepare_dict.sh /export/corpora5/TEDLIUM_release2
+local/cmu_tedlium_prepare_dict.sh $TEDLIUM_DIR
 
 dir=data/local/dict_combined
 swbd_dir=data/local/dict_swbd

egs/multi_en/s5/local/swbd1_data_download.sh

@@ -33,7 +33,7 @@ mkdir -p $dir
 
 # Audio data directory check
 if [ ! -d $SWBD_DIR ]; then
-  echo "Error: run.sh requires a directory argument"
+  echo "Error: swbd1_data_download.sh requires a valid directory argument"
   exit 1; 
 fi  
 

egs/multi_en/s5/local/swbd1_prepare_dict.sh

@@ -25,7 +25,9 @@ srcdict=$srcdir/swb_ms98_transcriptions/sw-ms98-dict.text
 # assume swbd_p1_data_prep.sh was done already.
 [ ! -f "$srcdict" ] && echo "$0: No such file $srcdict" && exit 1;
 
+rm -f $dir/lexicon0.txt
 cp $srcdict $dir/lexicon0.txt || exit 1;
+chmod +w $dir/lexicon0.txt
 patch <local/swbd_dict.patch $dir/lexicon0.txt || exit 1;
 
 #(2a) Dictionary preparation:

egs/multi_en/s5/run.sh

@@ -103,7 +103,7 @@ fi
 # across all training transcripts.
 if [ $stage -le 4 ]; then
   # We prepare the dictionary in data/local/dict_combined.
-  local/prepare_dict.sh
+  local/prepare_dict.sh $swbd $tedlium2
   local/g2p/train_g2p.sh --stage 0 --silence-phones "data/local/dict_combined/silence_phones.txt" data/local/dict_combined exp/g2p
   dict_dir=data/local/dict_nosp
   mkdir -p $dict_dir
@@ -148,19 +148,6 @@ fi
 
 # fix and validate training data directories
 if [ $stage -le 8 ]; then
-  # create segments file for wsj
-  awk '{print $1, $1, 0, -1}' data/wsj/train/utt2spk > data/wsj/train/segments
-  for f in `awk '{print $5}' data/wsj/train/wav.scp`; do
-    head -c 1024 $f | grep sample_count | awk '{print $3/16000}'
-  done > wsj_durations
-  paste -d' ' <(cut -d' ' -f1-3 data/wsj/train/segments) wsj_durations > wsj_segments
-  mv data/wsj/train/segments{,.bkp}
-  mv wsj_segments data/wsj/train/segments
-  rm -f wsj_segments wsj_durations
-  # create segments files for librispeech
-  for c in librispeech_100 librispeech_360 librispeech_500; do
-    awk '{print $1, $1, 0, $2}' data/$c/train/utt2dur > data/$c/train/segments;
-  done
   # get rid of spk2gender files because not all corpora have them
   rm -f data/*/train/spk2gender
   # create reco2channel_and_file files for wsj and librispeech