ASR:2014-12-29

From cslt Wiki
Jump to: navigation, search

Speech Processing

AM development

Environment

  • Modification
  • First down-frequency of gpu760
  • Improved the gpu Fan-speed
  • Change the sleep-mode of gpu
  • May gpu760 of grid-14 be something wrong. To be exchanged.
  • To buy 3*2k PCs.

Sparse DNN

RNN AM

A new nnet training scheduler

Dropout & Maxout & Convolutive network

  • Drop out(+)
    • Find and test unknown noise test-data.(++)
  • MaxOut && P-norm
  • Need to solve the too small learning-rate problem
    • Add one normalization layer after the pnorm-layer
    • Add L2-norm upper bound
  • Convolutive network(DAE)

DAE(Deep Atuo-Encode-DNN)

VAD

  • Harmonics and Teager energy features being investigation (++)

Speech rate training

Confidence

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

Speaker ID

  • Non-stream GMM:wer-2.28%
  seperate3-ivector:wer-3.54 single-ivector:wer-1.57  
  seperate-PLDA:wer-0.87 single-PLDA:wer-1.00   
  • Code ready

Language ID

Voice Conversion

  • Yiye is reading materials(+)


Text Processing

LM development

Domain specific LM

  • LM2.0
  • data check for lexicon(jietong)
  • merge lm with NAME POI etc.(hanzhenglong/wxx)
  • mix the sougou2T-lm,kn-discount continue
  • train a large lm using 25w-dict.(hanzhenglong/wxx)
  • prun history lm(wxx)
  • new dict.
  • dongxu help zhenglong with large dictionary.

tag LM

  • need to do
  • tag Probability should test add the weight(hanzhenglong) and handover to hanzhenglong (hold)
paper
  • modify the paper(yuanb two days),paper submit this week.

RNN LM

  • rnn
  • test wer RNNLM on Chinese data from jietong-data(this week)
  • generate the ngram model from rnnlm and test the ppl with different size txt.[1]
  • lstm+rnn
  • check the lstm-rnnlm code about how to Initialize and update learning rate.(hold)

Word2Vector

W2V based doc classification

  • Initial results variable Bayesian GMM obtained. Performance is not as good as the conventional GMM.(hold)
  • Non-linear inter-language transform: English-Spanish-Czch: wv model training done, transform model on investigation

Knowledge vector

  • Knowledge vector
  • Make a proper test set.
  • Modify the object function and training process.
  • Read Liu's paper.

relation

  • Accomplish transE with almost the same performance as the paper did(even better)[2]

Character to word

  • Character to word conversion(hold)
  • prepare the task: word similarity
  • prepare the dict.

Translation

  • v5.0 demo released
  • cut the dict and use new segment-tool

QA

improve fuzzy match

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

improve lucene search

  • add more feature to improve search.
  • POS, NER ,tf ,idf ..

XiaoI framework

  • context in xiaoI

query normalization

  • using NER to normalize the word
  • new inter will install SEMPRE