Sinovoice-2014-02-25 - 版本历史

2014年2月27日 (四) 07:12 Zhangzy

2014-02-27T07:12:24Z

2014-02-25T13:49:28Z

‎Training Analysis

2014-02-25T13:35:52Z

2014-02-25T10:34:26Z

‎Training Analysis

2014-02-25T10:33:46Z

2014-02-25T10:31:01Z

2014-02-25T10:30:02Z

‎Corpora

2014-02-25T07:51:54Z

‎470 + 300h + BJ mobile 105h training

2014-02-25T07:51:23Z

‎BJ mobile incremental training

2014-02-25T07:50:36Z

‎Auto Transcription

@@ 第50行： / 第50行： @@
 |large LM, it 5,  -5/-10 || 13.77 || 1.29
 |-
-|large LM, it 6, -5/-9    || 13.64 || -
+|large LM, it 6, -5/-9    || 13.64 || 1.12
 |-
-|large LM, it 6,  -5/-10 || 13.25 || -
+|large LM, it 6,  -5/-10 || 13.25 || 1.33
 |-
-|large LM, it 7, -5/-9    || 13.29 || -
+|large LM, it 7, -5/-9    || 13.29 || 1.12
 |-
-|large LM, it 7,  -5/-10 || 12.87 || -
+|large LM, it 7,  -5/-10 || 12.87 || 1.17
 |-
 |large LM, it 8, -5/-9    || 13.09 || -
@@ 第66行： / 第66行： @@
 |large LM, it 9,  -5/-10 || 12.55 || -
 |-
-|large LM, it 10, -5/-9    || 12.83 || -
+|large LM, it 10, -5/-9    || 12.83 || 1.51
 |-
-|large LM, it 10,  -5/-10 || 12.48 || -
+|large LM, it 10,  -5/-10 || 12.48 || 1.65
 |-
-|large LM, it 11, -5/-9    || 12.87 || -
+|large LM, it 11, -5/-9    || 12.87 || 1.61
 |-
-|large LM, it 11,  -5/-10 || 12.46 || -
+|large LM, it 11,  -5/-10 || 12.46 || 1.28
 |-
 |}

@@ 第89行： / 第89行： @@
 MPE2    14.22%      14.09%
 MPE3    14.26%      13.80%
-MPE4    14.24%
+MPE4    14.24%      13.68%
 </pre>

@@ 第87行： / 第87行： @@
 Xent    15.42%      14.78%
 MPE1    14.46%      14.23%
-MPE2    14.22%
+MPE2    14.22%      14.09%
-MPE3    14.26%
+MPE3    14.26%      13.80%
 </pre>

@@ 第82行： / 第82行： @@
 ===Training Analysis===
 * Shared tree GMM model training completed, WER% is similar to non-shared model .
-* Selected 100h online data, trained two systems: (1) di-syllable system (as the one used in the current training) (2) jt-phone system
+* Selected 100h online data, trained two systems: (1) di-syllable system (2) jt-phone system
 <pre>
          di-syl      jt-ph

@@ 第80行： / 第80行： @@
 * First version of DT model would be trained with online data (1700h)
-==Training Analysis===
+===Training Analysis===
 * Shared tree GMM model training completed, WER% is similar to non-shared model .
 * Selected 100h online data, trained two systems: (1) di-syllable system (as the one used in the current training) (2) jt-phone system

@@ 第14行： / 第14行： @@
 * 16k 6000h data: 978h online data from DataTang + 656h online mobile data + 4300h recording data
-==470 hour 8k training==
+==Telephone model training==
 ===470 + 300h + BJ mobile 105h training===

@@ 第9行： / 第9行： @@
 ==Corpora==
-* PICC data are under labeling (200h), ready in two weeks.
+* PICC data are under labeling (200h), ready in one week.
 * 105h data from BJ mobile
 * Now totally 1121h (470 + 346 + 105 + 200) telephone speech will be ready soon.
 * 16k 6000h data: 978h online data from DataTang + 656h online mobile data + 4300h recording data
 ==470 hour 8k training==

@@ 第19行： / 第19行： @@
 ===470 + 300h + BJ mobile 105h training===
 (1) 105 BJ mobile  re-training without NOISE: 33.97% WER
 (2) 105 BJ mobile  re-training with NOISE phone in training, but decoding without NOISE: 34.27%
 (3) (2) + noise-decoding (with noise phone in lexicon/LM), still under investigation

←上一版本		2014年2月25日 (二) 07:51的版本
第25行：		第25行：

	(1) Original 470 + 300 model: 30.24% WER		(1) Original 470 + 300 model: 30.24% WER
−	~~(2) Incremental DT training with 105h BJ data, 27.01% WER~~

		+	(2) Incremental DT training with 105h BJ data, 27.01% WER

	==6000 hour 16k training==		==6000 hour 16k training==

@@ 第95行： / 第95行： @@
 * PICC auto-trans incremental DT training completed
 Threshold  WER
 org:     45.03%
 .9:     41.89%
 .8:     41.64%
 * Current training data with 0.8 involve 80k sentences, amounting to about 60h data.
 * Sampling 60h labelled data to enrich the training
 =DNN Decoder=