HUB4 English Broadcast News recipe

egs/hub4_english/s5/README

@@ -0,0 +1,33 @@
+This is the English Broadcast News (HUB4) corpus.
+
+1996 English Broadcast News Train (HUB4)
+   Speech      LDC97S44
+   Transcripts LDC97T22
+
+1997 English Broadcast News Train (HUB4)
+  Speech       LDC98S71
+  Transcripts  LDC98T28
+
+1995 English Broadcast News (CSR-IV HUB4)
+  LDC96S31
+
+North American News Text Corpus
+  LDC95T21
+
+North American News Text Supplement Corpus
+  LDC98T30
+
+1996 CSR HUB4 Language Model
+  LDC98T31
+
+1996 English Broadcast News Dev and Eval (HUB4) 
+  LDC97S66
+
+1997 HUB4 English Evaluation corpus
+  LDC2002S11
+
+1998 HUB4 Broadcast News Evaluation English Test Material
+  LDC2000S86 
+
+1999 HUB4 Broadcast News Evaluation English Test Material
+  LDC2000S88

egs/hub4_english/s5/RESULTS

@@ -0,0 +1,9 @@
+for x in exp/*/decode*; do grep Sum $x/score*/*.ctm.*sys | utils/best_wer.sh ; done | sort -k2,2n
+exit 0
+
+%WER 17.8 | 728 32834 | 84.1 11.8 4.1 1.9 17.8 82.8 | exp/tri4/decode_nosp_eval97.pem_rescore/score_13_0.5/eval97.pem.ctm.filt.sys
+%WER 19.0 | 728 32834 | 83.0 12.7 4.3 2.0 19.0 84.2 | exp/tri4/decode_nosp_eval97.pem/score_13_0.0/eval97.pem.ctm.filt.sys
+%WER 19.4 | 728 32834 | 82.7 13.1 4.2 2.1 19.4 83.8 | exp/tri3/decode_nosp_eval97.pem_rescore/score_13_0.0/eval97.pem.ctm.filt.sys
+%WER 20.5 | 728 32834 | 81.7 13.9 4.4 2.3 20.5 85.0 | exp/tri3/decode_nosp_eval97.pem/score_13_0.0/eval97.pem.ctm.filt.sys
+%WER 23.7 | 728 32834 | 79.0 16.0 5.0 2.7 23.7 85.3 | exp/tri4/decode_nosp_eval97.pem.si/score_12_0.0/eval97.pem.ctm.filt.sys
+%WER 25.7 | 728 32834 | 77.1 17.6 5.3 2.8 25.7 85.9 | exp/tri3/decode_nosp_eval97.pem.si/score_13_0.0/eval97.pem.ctm.filt.sys

egs/hub4_english/s5/cmd.sh

@@ -0,0 +1,14 @@
+# you can change cmd.sh depending on what type of queue you are using.
+# If you have no queueing system and want to run on a local machine, you
+# can change all instances 'queue.pl' to run.pl (but be careful and run
+# commands one by one: most recipes will exhaust the memory on your
+# machine).  queue.pl works with GridEngine (qsub).  slurm.pl works
+# with slurm.  Different queues are configured differently, with different
+# queue names and different ways of specifying things like memory;
+# to account for these differences you can create and edit the file
+# conf/queue.conf to match your queue's configuration.  Search for
+# conf/queue.conf in http://kaldi-asr.org/doc/queue.html for more information,
+# or search for the string 'default_config' in utils/queue.pl or utils/slurm.pl.
+
+export train_cmd="queue.pl --mem 1G"
+export decode_cmd="queue.pl --mem 4G"

egs/hub4_english/s5/conf/mfcc.conf

@@ -0,0 +1 @@
+--use-energy=false   # only non-default option.

egs/hub4_english/s5/conf/vad.conf

@@ -0,0 +1,2 @@
+--vad-energy-threshold=5.5
+--vad-energy-mean-scale=0.5