“ASR:2015-03-09”版本间的差异
来自cslt Wiki
（→Text Processing） |
|||
| 第57行： | 第57行： | ||
====Domain specific LM==== | ====Domain specific LM==== | ||
* LM2.X | * LM2.X | ||
| − | |||
:* train a large lm using 25w-dict.(hanzhenglong/wxx) | :* train a large lm using 25w-dict.(hanzhenglong/wxx) | ||
::* v2.0c filter the useless word.(next week) | ::* v2.0c filter the useless word.(next week) | ||
| 第71行： | 第70行： | ||
====RNN LM==== | ====RNN LM==== | ||
*rnn | *rnn | ||
| − | :* | + | :* the input and output is word embedding and add some token information like NER.. |
| + | :* map the word to character and train the lm. | ||
*lstm+rnn | *lstm+rnn | ||
:* check the lstm-rnnlm code about how to Initialize and update learning rate.(hold) | :* check the lstm-rnnlm code about how to Initialize and update learning rate.(hold) | ||
| 第89行： | 第89行： | ||
===Sparse NN in NLP=== | ===Sparse NN in NLP=== | ||
* prepare the ACL | * prepare the ACL | ||
| − | + | :* check the code to find the problem . | |
| + | :* increase the dimension | ||
| + | :* use different test set. | ||
===QA=== | ===QA=== | ||
====improve fuzzy match==== | ====improve fuzzy match==== | ||
| 第95行： | 第97行： | ||
===online learning=== | ===online learning=== | ||
* data is ready.prepare the ACL paper | * data is ready.prepare the ACL paper | ||
| − | ==== | + | :* prepare sougouQ data and test it using current online learning method |
| − | * | + | ====framework==== |
| − | :* | + | * extract the module |
| − | + | :* extract the context module ,search module,entity recognize module and common module. | |
| + | :* define the inference in different modules | ||
| + | * composite module | ||
| + | ====leftover problem==== | ||
* new inter will install SEMPRE | * new inter will install SEMPRE | ||
2015年3月16日 (一) 01:11的最后版本
Speech Processing
AM development
Environment
- grid-11 often shut down automatically, too slow computation speed.
- GPU has being repired.--Xuewei
RNN AM
- details at http://liuc.cslt.org/pages/rnnam.html
- triphone one state based RNN ?--Liu Chao
Mic-Array
- reproduce environment for interspeech
- alpha parameter in Lasso
Dropout & Maxout & rectifier
- HOLD
- Need to solve the too small learning-rate problem
- 20h small scale sparse dnn with rectifier. --Mengyuan
- 20h small scale sparse dnn with Maxout/rectifier based on weight-magnitude-pruning. --Mengyuan Zhao
Convolutive network
- Convolutive network(DAE)
- HOLD
- Technical report writing, Mian Wang, Yiye Lin, Shi Yin, Mengyuan Zhao
- reproduce experiments -- Yiye
DNN-DAE(Deep Auto-Encode-DNN)
- HOLD
- Technical report to draft, Xiangyu Zeng, Shi Yin, Mengyuan Zhao and Zhiyong Zhang,
- http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=318
RNN-DAE(Deep based Auto-Encode-RNN)
- HOLD -Zhiyong
- http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=261
Speech rate training
- http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=268
- Technical report to draft.-- Xiangyu Zeng, Shi Yin
- Prepare for NCMSSC
Neural network visulization
- http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=324
- Technical report writing --Mian Wang.
Speaker ID
Ivector based ASR
- Ivector dimention is smaller, performance is better
- Augument to hidden layer is better than input layer
Text Processing
LM development
Domain specific LM
- LM2.X
- train a large lm using 25w-dict.(hanzhenglong/wxx)
- v2.0c filter the useless word.(next week)
- set the test set for new word (hold)
- prepare the wiki data: entity list.
tag LM
- Tag Lm(JT)
- error check
- similar word extension in FST
- add the experiment to tag-lm paper.
RNN LM
- rnn
- the input and output is word embedding and add some token information like NER..
- map the word to character and train the lm.
- lstm+rnn
- check the lstm-rnnlm code about how to Initialize and update learning rate.(hold)
Word2Vector
W2V based doc classification
- data prepare.(hold)
Knowledge vector
- make a report on Monday
Translation
- v5.0 demo released
- cut the dict and use new segment-tool
Sparse NN in NLP
- prepare the ACL
- check the code to find the problem .
- increase the dimension
- use different test set.
QA
improve fuzzy match
- add Synonyms similarity using MERT-4 method(hold)
online learning
- data is ready.prepare the ACL paper
- prepare sougouQ data and test it using current online learning method
framework
- extract the module
- extract the context module ,search module,entity recognize module and common module.
- define the inference in different modules
- composite module
leftover problem
- new inter will install SEMPRE
