Edinburgh Speech Tools  2.1-release
wfst_run

Run a weighted finite-state transducer

Synopsis

wfst_run [WFSTFILE] [input file0] ... [-o output file] [-wfst ifile] [-transduce ] [-recog ] [-cumulate_into ofile] [-itype string] [-quiet ] [-perplexity ] [-heap int] [-o ofile]

This program runs a WFST on some given data. It works in either recognize mode where both inputs and output are specified, but also in transduction mode where an input is transduced to the output.

Options

  • -wfst: ifile The WFST to use
  • -transduce: Transduce input to output (default)
  • -recog: Recognize input consists of pairs
  • -cumulate_into: ofile Cumulate transitions to give new weights save new WFST into ofile
  • -itype: string char or token
  • -quiet: No extraneous messages
  • -perplexity: Calculate perplexity on given data set
  • -heap: int " {210000}" Set size of Lisp heap, needed for large wfsts
  • -o: ofile Output file for transduced forms