ASR:2014-12-15

来自cslt Wiki
跳转至： 导航、 搜索

Speech Processing

AM development

Environment

  • Already buy 3 760GPU
  • grid-9/12 760GPU crashed again; grid-11 shutdown automatically.
  • Change 760gpu card of grid-12 and grid-14(+).
  • First down-frequency of gpu760.

Sparse DNN

RNN AM

  • Initial nnet seems not very well, need to be pre-trained or test lower learn-rate.
  • For AURORA 4 1h/epoch, model train done.
  • Using AURORA 4 short-sentence with a smaller number of targets.(+)
  • Adjusting the learning rate.(+)
  • Trying toolkit of Microsoft.(+)
  • details at http://liuc.cslt.org/pages/rnnam.html
  • Reading papers

A new nnet training scheduler

Dropout & Maxout & Convolutive network

  • Drop out(+)
  • Use different proportion of noise data to investigate the effect of xEnt and mpe and dropout
    • Problem 1) The effect of dropout in different noise proportion; 
           2) The effect of MPE in different noise proportion;
           3) The effect of MPE+dropout in different noise proportion.

 Dropout is effective for minority.
    • Find and test unknown noise test-data.(++)
  • MaxOut
  • Pretraining based maxout, can't use large learning-rate.
  • 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(Deep Atuo-Encode)

 (1) train_clean
   drop-retention/testcase(WER)| test_clean_wv1  | test_airport_wv1 | test_babble_wv1 | test_car_wv1 
  ---------------------------------------------------------------------------------------------------------
      std-xEnt-sigmoid-baseline| 6.04            |    29.91         |   27.76         | 16.37
  ---------------------------------------------------------------------------------------------------------
      std+dae_cmvn_noFT_2-1200 | 7.10            |    15.33         |   16.58         | 9.23
  ---------------------------------------------------------------------------------------------------------
   std+dae_cmvn_splice5_2-100  | 8.19            |    15.21         |   15.25         | 9.31
  ---------------------------------------------------------------------------------------------------------

Denoising & Farfield ASR

  • ICASSP paper submitted.
  • HOLD

VAD

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

Speech rate training

  • Data ready on tencent set; some errors on speech rate dependent model. error fixed.
  • Retrain new model(+)

Scoring

  • Timber Comparison done.
  • harmonics based timber comparison: frequency based feature is better. done
  • GMM based timber comparison is done. Similar to speaker recognition. done
  • TODO: Code checkin and technique report. done

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

  • Preparing GMM-based server.
  • EER ~ 4% (GMM-based system)--Text independent
  • EER ~ 6%(1s) / 0.5%(5s) (GMM-based system)--Text dependent
  • test different number of components; fast i-vector computing
  • Test with number recordings, The 256 number component is the best.
  • Test with text-dependent recordings, The 1024 number component is the best.
  • Interpolation alpha is not sensitive.

Language ID

  • GMM-based language is ready.
  • Delivered to Jietong
  • Prepare the test-case

Voice Conversion

  • Yiye is reading materials(+)


Text Processing

LM development

Domain specific LM

  • domain lm
  • Sougou2T : kn-count continue .
  • lm v2.0.
  • problem that the result is different between kaldi and jitong.
  • check the format of data encode in kaldi and jitong.

tag LM

  • paper(two week)
  • add more related work in introduction and rich the paper

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

  • baseline
  • Debug the training part and add normalization, result in cvss[326]. 
  Category:Mammal orders
    entry number: 15572
    testset: 70 couples (28 data scored 0, 11 data scored 1, 16 data scored 2,15 data scored 3)
                   Spearman rank-order coefficient | Pearson correlation coefficient
Factorization:      0.698714                          0.747226        
Baseline:           0.721214                          0.778838
    testset: 70 couples (7 data scored 0, 2 data scored 1, 5 data scored 2, 8 data scored 3)
Baseline:(category) 0.523433
  • Finish the baseline and compared with simple factorization method based on co-occurrence.
  • Visualization of knowledge vector[2]
  • test
  • set up the test data.

relation

  • Accomplish transE with almost the same performance as the paper did(even better)[3]

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

  • domain keyword to improve search in lucene using mert
  • keyword in sentence,like frequency ,position.
  • keyword like pos,parser information.

XiaoI framework

  • context in QA

query normalization

  • using NER to normalize the word
  • new inter will install SEMPRE
取自“http://index.cslt.org/mediawiki/index.php?title=ASR:2014-12-15&oldid=13086”