1 Reply Latest reply: May 23, 2012 11:05 AM by Roger Ford-Oracle RSS

    Speeding up SVN_CLASSIFIER training

    939239
      I'm relatively new to Oracle Text, so I'm quite possibly missing something obvious here, but is there a means of forcing parallelism on the CTX_CLS.TRAIN procedure or otherwise optimizing the process?

      I'm setting things up much as per the example in the Oracle Text dev guide.

      I've tried setting CREATE INDEX with a PARALLEL clause, but as it contains the 'nopopulate' preference my guess is that this doesn't make a difference?

      I've also explored whether it's possible to set exec ctx_cls.train(...) with a PARALLEL clause but nothing seems suggest it should be possible.

      Any thoughts? My sample application has a training set of about 7,000 records across three categories; that doesn't feel like a lot but maybe it is?