Dataset Open Access

NAIST contribution to ZeroSpeech 2017 (Track 1)

Heck, Michael; Sakti, Sakriani; Nakamura, Satoshi


JSON-LD (schema.org) Export

{
  "@context": "https://schema.org/", 
  "@id": "https://doi.org/10.5281/zenodo.821246", 
  "@type": "Dataset", 
  "creator": [
    {
      "@type": "Person", 
      "affiliation": "NAIST", 
      "name": "Heck, Michael"
    }, 
    {
      "@type": "Person", 
      "affiliation": "NAIST", 
      "name": "Sakti, Sakriani"
    }, 
    {
      "@type": "Person", 
      "affiliation": "NAIST", 
      "name": "Nakamura, Satoshi"
    }
  ], 
  "datePublished": "2017-06-30", 
  "description": "<p>This is the official submission of NAIST for track 1 of the zero resource speech challenge 2017 (ZeroSpeech2017).</p>\n\n<p>Our system uses feature vector optimized DPGMM based clustering for unsupervised subword modeling. The general idea is to unsupervisedly learn frame-level class labels in a first run of DPGMM based clustering. These labels are then used to automatically learn speech feature transformations (LDA, MLLT, (basis) fMLLR)\u00a0to improve discriminability and to reduce speaker variance. The optimized feature vectors are\u00a0re-clustered and frame-wise posteriorgrams are extracted to serve as new speech representation.</p>\n\n<p>The system also applies\u00a0posteriorgram based model combination.</p>\n\n<p>This pipeline is entirely unsupervised and only needs the raw audio recordings as input. No pre-defined segmentation, speaker IDs or other meta data is required.</p>\n\n<p>The only parameter subject to tuning is the LDA output dimensionality, which in this case has been optimized on the development data sets (english, french, mandarin) and tested on the surprise language data sets (LANG1, LANG2).</p>\n\n<p>Since the output is posteriorgrams, the ABX scoring needs to be done with the Kullback-Leibler (KL) divergence.</p>", 
  "distribution": [
    {
      "@type": "DataDownload", 
      "contentUrl": "https://zenodo.org/api/files/e05c9ae1-2641-4dab-b06f-4d1aeb77d5b1/10_5281_zenodo_821246.tar.gz", 
      "fileFormat": "gz"
    }
  ], 
  "identifier": "https://doi.org/10.5281/zenodo.821246", 
  "keywords": [
    "zerospeech2017, dpgmm"
  ], 
  "license": "https://creativecommons.org/licenses/by/4.0/", 
  "name": "NAIST contribution to ZeroSpeech 2017 (Track 1)", 
  "url": "https://zenodo.org/record/821246"
}
52
9
views
downloads
All versions This version
Views 5252
Downloads 99
Data volume 80.9 GB80.9 GB
Unique views 5252
Unique downloads 88

Share

Cite as