“ASR:2015-03-02”版本间的差异

2015年3月6日 (五) 07:25的版本

Speech Processing

AM development

Environment

• grid-11 often shutdown automatically, too slow computation speed.
• buy a new 800W power -- Xuewei

RNN AM

• details at http://liuc.cslt.org/pages/rnnam.html
• triphone one state based RNN?

Mic-Array

• the technical report is done.
• reproduce environment for interspeech

Dropout & Maxout & rectifier

• HOLD
• Need to solve the too small learning-rate problem
• 20h small scale sparse dnn with rectifier. --Chao liu
• 20h small scale sparse dnn with Maxout/rectifier based on weight-magnitude-pruning. --Mengyuan Zhao

Convolutive network

• Convolutive network(DAE)

• http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=311
• 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
• http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=261

VAD

• DAE
• Technical report done. -- Shi Yin

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

Confidence

• HOLD
• Reproduce the experiments on fisher dataset.
• Use the fisher DNN model to decode all-wsj dataset
• preparing scoring for puqiang data

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

• DNN-based sid
• http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=327

Text Processing

LM development

Domain specific LM

• LM2.X

• mix the sougou2T-lm,kn-discount(done)
• train a large lm using 25w-dict.(hanzhenglong/wxx)

• v2.0a adjust the weight and smaller weight of transcription is better.(done)
• v2.0b add the v1.0 vocab(done)
• v2.0c filter the useless word.(next week)
• set the test set for new word (hold)

tag LM

• Tag Lm

• add 3-class tag and test

• similar word extension in FST

• improve the key-word weight in G , and result is good in keyword recognization
• read to deal with the English-Chinese

RNN LM

• rnn

• test wer RNNLM on Chinese data from jietong-data
• generate the ngram model from rnnlm and test the ppl with different size txt.

• lstm+rnn

• check the lstm-rnnlm code about how to Initialize and update learning rate.(hold)

Word2Vector

W2V based doc classification

• data prepare.

Knowledge vector

• paper is done, submitted ACL

Character to word

• Character to word conversion(hold)

Word vector online learing

• prepare the ACL

Translation

• v5.0 demo released

• cut the dict and use new segment-tool

Sparse NN in NLP

• write a technical report(Wednesday) and make a report.
• ready to prepare the ACL

QA

improve fuzzy match

• add Synonyms similarity using MERT-4 method(hold)

improve lucene search

• commit to Rong Liu to check in.

• online learning to rank

• tool is ok and ready to test

online learning

• a simple edition about online learning part about QA.

context framework

• code for organization

• change to knowledge graph,and learn the D2R tool and JENA 政府组织推演 员工信息实例推演
• code for demo

query normalization

• using NER to normalize the word

• new inter will install SEMPRE