cslt Wiki - 用户贡献 [zh-cn]

Hulan-2013-10-18

2013-10-18T02:58:17Z

Litc：

=ASR=

==ASR Kernel development==

[[http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/2013-10-18 ASR group weekly report]]

==TTS==

* full-lab training is ready. Trained the first full-lab system with 16k/pseduo 48k data.
* re-recording 48k data using F00 (500 sentences) and retrain the model. The quality of the signal sounds better, while the quality of pitch is a bit strange. Need more investigation on parameter settings.

Next week:

* Check the signal parameters and solve the problem of pitch.
* Prepare the large data training with both all-F 863 data.
* Prepare the large data training with online novel.

=Dialog system=

* The search system migrated to the custom domain, with significant performance reduction

<pre>
Customs:
n TF TFIDF
1 0.496 0.485
2 0.619 0.615
3 0.676 0.673
4 0.713 0.715
5 0.740 0.738

Agriculture:
n TF TFIDF
1 0.75 0.8
2 0.85 0.883
3 0.867 0.917
4 0.867 0.95
5 0.95 0.967
</pre>

* Two problems:
# short of semantic cluster.
# limited training data for idf.

* Next week
# Analyse the QA database, to extract useful domain dependent data
# Analyse the data to expand the key words & phrases
# Analyse the data to attain better IDF.

=Summary system=

* Be familiar with the dragon system. Combing the system and extract the summary-only code.
* Sentence based summary done. But request to migrate to Chinese.
* Start to build the textrank-based keyword extraction. Re-write the Lexrank code to handle word level similarity matrices.

Next week:

* Test data set: 100 articles
* TextRank Done

==Template matching==

* Start to work on the self coding, while some requests have not been considered.
* consider if to use the standard FSM toolkit by next Tuesday.

Hulan-2013-10-18

2013-10-18T01:58:21Z

Litc：/* TTS */

Hulan-2013-10-18

2013-10-18T01:40:25Z

Litc：/* ASR Kernel development */

=ASR=

==ASR Kernel development==

[[http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/2013-10-18 ASR group weekly report]]

==TTS==

* CD lab files done. Refining the script.
* Training toolkit is cleaned up. Now no alignment is required. Parallel training is done.
* Tried syllable based system instead of phones.
* Collected an online-novel reading.

Next week:

* Refine the script
* Clean up the online reading.

=Dialog system=

* The search system migrated to the custom domain, with significant performance reduction

<pre>
Customs:
n TF TFIDF
1 0.496 0.485
2 0.619 0.615
3 0.676 0.673
4 0.713 0.715
5 0.740 0.738

Agriculture:
n TF TFIDF
1 0.75 0.8
2 0.85 0.883
3 0.867 0.917
4 0.867 0.95
5 0.95 0.967
</pre>

* Two problems:
# short of semantic cluster.
# limited training data for idf.

* Next week
# Analyse the QA database, to extract useful domain dependent data
# Analyse the data to expand the key words & phrases
# Analyse the data to attain better IDF.

Hulan-2013-10-18

2013-10-18T01:40:11Z

Litc：/* Dialog system */

=ASR=

==ASR Kernel development==

[[http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/2013-10-11 ASR group weekly report]]

==TTS==

* CD lab files done. Refining the script.
* Training toolkit is cleaned up. Now no alignment is required. Parallel training is done.
* Tried syllable based system instead of phones.
* Collected an online-novel reading.

Next week:

* Refine the script
* Clean up the online reading.

=Dialog system=

* The search system migrated to the custom domain, with significant performance reduction

<pre>
Customs:
n TF TFIDF
1 0.496 0.485
2 0.619 0.615
3 0.676 0.673
4 0.713 0.715
5 0.740 0.738

Agriculture:
n TF TFIDF
1 0.75 0.8
2 0.85 0.883
3 0.867 0.917
4 0.867 0.95
5 0.95 0.967
</pre>

* Two problems:
# short of semantic cluster.
# limited training data for idf.

* Next week
# Analyse the QA database, to extract useful domain dependent data
# Analyse the data to expand the key words & phrases
# Analyse the data to attain better IDF.

Hulan-2013-10-18

2013-10-18T01:39:52Z

Litc：以内容“=ASR= ==ASR Kernel development== http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/2013-10-11 ASR group weekly report ==TTS== * CD lab files done. Refining ...”创建新页面

=ASR=

==ASR Kernel development==

[[http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/2013-10-11 ASR group weekly report]]

==TTS==

* CD lab files done. Refining the script.
* Training toolkit is cleaned up. Now no alignment is required. Parallel training is done.
* Tried syllable based system instead of phones.
* Collected an online-novel reading.

Next week:

* Refine the script
* Clean up the online reading.

=Dialog system=

* The search system migrated to the custom domain, with significant performance reduction

Customs:
n TF TFIDF
1 0.496 0.485
2 0.619 0.615
3 0.676 0.673
4 0.713 0.715
5 0.740 0.738

Agriculture:
n TF TFIDF
1 0.75 0.8
2 0.85 0.883
3 0.867 0.917
4 0.867 0.95
5 0.95 0.967

* Two problems:
# short of semantic cluster.
# limited training data for idf.

* Next week
# Analyse the QA database, to extract useful domain dependent data
# Analyse the data to expand the key words & phrases
# Analyse the data to attain better IDF.

Weekly status

2013-10-18T01:27:41Z

Litc：

2013-10-18

2013-10-18T01:26:59Z

Litc：/* Noisy training */

== Data sharing ==

* LM count files still undelivered!

== DNN progress ==

=== Sparse DNN ===

* Optimal Brain Damage(OBD). Code ready, bug found.

=== Tencent exps ===
N/A

==Noisy training ==

1. With 863 clean test, by adding car & white noise at various levels, obtained significant performance improvement.

2. The test with both car & white noise benefits from the noisy training.

==Continuous LM ==

1. Lattice rescoring toolkit is ready.
2. Rescoring is slow with some dense lattices.

==QA LM==

1. use the QA word segment system
2. train the Q LM & QA ASR system

2013-10-18

2013-10-18T01:26:48Z

Litc：以内容“== Data sharing == * LM count files still undelivered! == DNN progress == === Sparse DNN === * Optimal Brain Damage(OBD). Code ready, bug found. === Tencent exps ...”创建新页面

== Data sharing ==

* LM count files still undelivered!

== DNN progress ==

=== Sparse DNN ===

* Optimal Brain Damage(OBD). Code ready, bug found.

=== Tencent exps ===
N/A

==Noisy training ==

1. With 863 clean test, by adding car & white noise at various levels, obtained significant performance improvement.
2. The test with both car & white noise benefits from the noisy training.

==Continuous LM ==

1. Lattice rescoring toolkit is ready.
2. Rescoring is slow with some dense lattices.

==QA LM==

1. use the QA word segment system
2. train the Q LM & QA ASR system

CSLT cluster nodes

2013-10-18T01:17:05Z

Litc：

'''警告：'''“CSLT cluster nodes”指向这里，但您没有足够的权限来访问它。

Use CSLT cluster

2013-10-18T01:15:22Z

Litc：/* Login and run jobs */

'''警告：'''“Use CSLT cluster”指向这里，但您没有足够的权限来访问它。