1 / 14

BUT SWS 2013 - Massive parallel approach

BUT SWS 2013 - Massive parallel approach Brno University of Technology Faculty of Information Technology Speech@FIT Igor Sz öke , Lukáš Burget, František Grézl , Lucas Ondel. MediaEval SWS 2013 workshop, October 18.-19. 2013, Barcelona. Outlines.

miette
Download Presentation

BUT SWS 2013 - Massive parallel approach

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. BUT SWS 2013 - Massive parallel approach Brno University of Technology Faculty of Information Technology Speech@FIT Igor Szöke, Lukáš Burget, František Grézl, Lucas Ondel MediaEval SWS 2013 workshop, October 18.-19. 2013, Barcelona

  2. Outlines • Systems overview & Underlyingtechnologies • AKWS • DTW • Calibration • Fusion • Resultsand discussion

  3. System overview • Ourinternaltaskwas: • To reuse as many Atomic systems as we have and • fuse them on the detection level. • We end up with: • 13 Atomic systems, 26 QbE sub-systems, • 19 languages (16 unique). • zero resourced system • Ingredients • Phonemerecognizer, AcousticKeywordSpotting, • DTW, Calibration, Fusion

  4. System overview Igor’s Greeting 

  5. Subsystem • Sentence meannormalization • Neuralnetwork basedfeatures • threestatephoneposteriors • Querydetector • AKWS • DTW

  6. Atomic system • Adaptation on target data (GP and BABEL NNs) • Original NN used for target data labeling (state level) • Then, universal context, bottle-neck neural network base • classifier trained. • LCRC, SWS2012 without any adaptation.

  7. AKWS QbE subsystem • Query -> example-to-text using phoneme recognizer • Omit initial and final silence • Omit queries having less than 3 non-silence phonemes • No LM constrains

  8. DTW QbE subsystem • Segmental DTW (query can start in any frame of utterance) • Log dot product over phoneme state posteriors • Path cost: 1, 1, 1 • On-line normalizing of the path • While filling a cell in a distant matrix, the value already • considers the length of the previous path • We add VAD as late submission -> really huge impact • Initial and final silence frames were removed from • examples

  9. Calibration • Really important! • No-norm, z-norm, z-norm_sideinfo, m-norm (the best) • Experiments with adding sideinfo [log(#term_occ), #phn, • log(#nonsilence frames)] • Linear model was trained (using logistic regresion) • Good improvement • M-norm – find the peak in histogram of term scores • Calculate variance of data <peak, +inf> • Apply variance norm on the whole data set • Subtract the peak (shift the peak to 0) • Event better than z-norm • Sideinfo does not helped! • (means m-norm is calibrated enough)

  10. DTW AWKS Orig Z-norm M-norm

  11. Calibration 1 AKWS subsystem MTWV (UBTWV) orig 0.0000 (0.1012) z-norm 0.0330 (0.1434) z-norm_side 0.0603 (0.1436) m-norm 0.0769 (0.1611)

  12. Fusion • Linear combination of subsystems (and one bias) • Trained with respect to minimizing of cross entropy • (binary logistic regression) • Detections are clustered • System not producing any score at given time get a • default score

  13. Fusion

  14. Results • MTWV(UBTWV) • UBTWV – non-pooled TWV, ideal calibration, oracle calibration • DTW is superior to AKWS… but the speed… • Still having some gaps in calibration • (the difference between DEV and EVAL TWV) • NN unsupervised adaptation helped • 1 AKWS subsystem: 0.0443(0.1154) -> 0.0769(0.1630) • m-norm! • Lot of directions for research

More Related