Commit 66cba1f2 authored by Gaurav Kumar's avatar Gaurav Kumar
Browse files

Trunk:fisher-callhome-spanish:Updated the RESULTS file with DNN results

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4596 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8
parent dfdfba6e
Kaldi recipe for the Fisher and Callhome Spanish Corpora
About the Fisher Spanish Corpus
Fisher Spanish - Speech was developed by the Linguistic
Data Consortium (LDC) and consists of audio files covering
roughly 163 hours of telephone speech from 136 native
Caribbean Spanish and non-Caribbean Spanish speakers.
Full orthographic transcripts of these audio files are available
in LDC2010T04
Speech : LDC2010S01
Transcripts : LDC2010T04
About the Callhome Spanish Corpus
The CALLHOME Spanish corpus of telephone speech consists
of 120 unscripted telephone conversations between native speakers of Spanish.
All calls, which lasted up to 30 minutes, originated in North America
and were placed to international locations. Most participants called
family members or close friends.
Speech : LDC96S35
Transcripts : LDC96T17
The LDC Spanish rule based lexicon
The CALLHOME Spanish collection includes a lexical component.
The CALLHOME Spanish Lexicon consists of 45,582 words and contains
separate information fields with phonological, morphological and
frequency information for each word.
Lexicon : LDC96L16
Each subdirectory of this directory contains the
scripts for a sequence of experiments.
s5: This recipe is based on the WSJ s5 recipe. It works with the
the transcripts (available along with the script in LDC97T19). In addition,
it uses a phonetic lexicon generated using the rules based LDC lexicon.
The recipe follows the Triphone+SGMM+SAT+fMLLR+SGMM+DNN pipeline. It uses data
partitions as specified by LDC in the Callhome corpus description. For Fisher
custom partitions are available (check the run.sh file for the location
of the split file : This can be changed).
......@@ -94,4 +94,17 @@ exp/sgmm5/decode_dev/wer_15:%WER 33.71 [ 13880 / 41177, 1709 ins, 4962 del, 7209
exp/sgmm5/decode_dev/wer_16:%WER 34.09 [ 14037 / 41177, 1602 ins, 5226 del, 7209 sub ]
exp/sgmm5/decode_dev/wer_8:%WER 34.04 [ 14016 / 41177, 3118 ins, 3059 del, 7839 sub ]
exp/sgmm5/decode_dev/wer_9:%WER 33.20 [ 13671 / 41177, 2807 ins, 3267 del, 7597 sub ]
--------------------------------------------------------------------------------------
pNorm-Ensemble DNN
--------------------------------------------------------------------------------------
exp/tri6a_dnn/decode_dev/wer_10:%WER 31.02 [ 12774 / 41177, 2762 ins, 3042 del, 6970 sub ]
exp/tri6a_dnn/decode_dev/wer_11:%WER 30.49 [ 12556 / 41177, 2573 ins, 3236 del, 6747 sub ]
exp/tri6a_dnn/decode_dev/wer_12:%WER 30.15 [ 12414 / 41177, 2384 ins, 3414 del, 6616 sub ]
exp/tri6a_dnn/decode_dev/wer_13:%WER 29.93 [ 12324 / 41177, 2237 ins, 3593 del, 6494 sub ]
exp/tri6a_dnn/decode_dev/wer_14:%WER 29.87 [ 12298 / 41177, 2093 ins, 3794 del, 6411 sub ]
exp/tri6a_dnn/decode_dev/wer_15:%WER 29.80 [ 12269 / 41177, 1946 ins, 3967 del, 6356 sub ]
exp/tri6a_dnn/decode_dev/wer_16:%WER 29.96 [ 12336 / 41177, 1869 ins, 4165 del, 6302 sub ]
exp/tri6a_dnn/decode_dev/wer_8:%WER 32.81 [ 13511 / 41177, 3317 ins, 2733 del, 7461 sub ]
exp/tri6a_dnn/decode_dev/wer_9:%WER 31.74 [ 13068 / 41177, 3017 ins, 2889 del, 7162 sub ]
--------------------------------------------------------------------------------------
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment