XML Output for Sphinx by bns26590


									      XML Output for Sphinx
• Motivation: applications may be able to
  make use of richer information from sphinx
  including n-best lists, the word lattice, and
  other features. An xml dtd format will be
  standard, and easy to parse, express, and
             Proposed DTD
– http://www.cs.cmu.edu/~tkharris/usi/utterance-0.1.dtd
– Sphinx produces utterances, each utterance is an xml
  document that conforms to the DTD
– An utterance is an n-best list or word-lattice or both
– An n-best list is a list of lists of words
– Each list and the words may have features
– The DTD desperately needs review

•   Is the motivation justified?
•   Computational/Network impact too much?
•   API’s are needed to parse XML
•   Need to get requirements/observations from
    Sphinx customers

To top