Skip to content

Instantly share code, notes, and snippets.

@renaud
Last active December 28, 2015 13:09
Show Gist options
  • Save renaud/7505848 to your computer and use it in GitHub Desktop.
Save renaud/7505848 to your computer and use it in GitHub Desktop.
DkPro Binary Cas evaluation

DkPro Binary Cas evaluation

Quick evaluation on 126 pdfs, with some preprocessing.

Findings:

Speed

just preprocessing (no BinaryCasWriter)

  114283ms   59.44%  Process            ch.epfl.bbp.uima.pdf.cr.PdfCollectionReader
    4505ms    2.34%  Analysis           ch.epfl.bbp.uima.ae.TokenAnnotator
   51355ms   26.71%  Analysis           ch.epfl.bbp.uima.ae.PosTagAnnotator
   21752ms   11.31%  Analysis           ch.epfl.bbp.uima.ae.BlueBioLemmatizer
  192263ms   TOTAL TIME (3 minutes and 12 seconds)

preprocessing and BinaryCasWriter, format: 0

  112566ms   14.23%  Process            ch.epfl.bbp.uima.pdf.cr.PdfCollectionReader
    4922ms    0.62%  Analysis           ch.epfl.bbp.uima.ae.TokenAnnotator
   50227ms    6.35%  Analysis           ch.epfl.bbp.uima.ae.PosTagAnnotator
   22071ms    2.79%  Analysis           ch.epfl.bbp.uima.ae.BlueBioLemmatizer
  601061ms   75.96%  Analysis           de.tudarmstadt.ukp.dkpro.core.io.bincas.BinaryCasWriter
  791257ms   TOTAL TIME (13 minutes and 11 seconds)

preprocessing and BinaryCasWriter with compression, format: 4

  110638ms   54.83%  Process            ch.epfl.bbp.uima.pdf.cr.PdfCollectionReader
    5394ms    2.67%  Analysis           ch.epfl.bbp.uima.ae.TokenAnnotator
   52900ms   26.22%  Analysis           ch.epfl.bbp.uima.ae.PosTagAnnotator
   22734ms   11.27%  Analysis           ch.epfl.bbp.uima.ae.BlueBioLemmatizer
    9670ms    4.79%  Analysis           de.tudarmstadt.ukp.dkpro.core.io.bincas.BinaryCasWriter
  201777ms   TOTAL TIME (3 minutes and 21 seconds)

reading BinaryCasReader

    3600ms   83.92%  Process            de.tudarmstadt.ukp.dkpro.core.io.bincas.BinaryCasReader
     660ms   15.38%  Analysis           ch.epfl.bbp.uima.ae.AddHeaderFromDkproMetadata
      24ms    0.56%  Analysis           ch.epfl.bbp.uima.ae.StatsAnnotatorPlus
    4290ms   TOTAL TIME (4 seconds)

reading BinaryCasReader, compressed

    2407ms   77.15%  Process            de.tudarmstadt.ukp.dkpro.core.io.bincas.BinaryCasReader
     663ms   21.25%  Analysis           ch.epfl.bbp.uima.ae.AddHeaderFromDkproMetadata
      41ms    1.31%  Analysis           ch.epfl.bbp.uima.ae.StatsAnnotatorPlus
    3120ms   TOTAL TIME (3 seconds)

Size

126 pdfs, choosen at random
pdfs: 391 MB
serialized: 358 MB
serialized compressed: 17 MB
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment