Some initial notes on the English paraphrase grammar

• There are three .cfg files in the scripts directory:
  • paraphrase.cfg. Use this to compile paraphrase grammar for generation (original purpose)
  • paraphrase_recogniser.cfg. Use this to compile paraphrase grammar for recognition
  • paraphrase_recogniser_parser.cfg. Use this to do speech recognition + parsing with paraphrase grammar
• The grammar files
  • The generation and recognition versions of the grammar are slightly different.
  • It's done using include files, with most of the rules shared.
  • Look in the .cfg files to see the details
• Make targets, in scripts/Makefile
  • paraphrase_grammar. Compile paraphrase grammar for generation
  • paraphrase_recogniser. Compile paraphrase grammar into recognition package
• Running the paraphrase recogniser
  • Start Regulus with paraphrase_recogniser_parser.cfg
  • LOAD to load normal paraphrase grammar for parsing
  • LOAD_RECOGNITION to start recogniser. This assumes a script in $REGULUS/scripts/run_license.bat which starts a license manager.
  • RECOGNISE to recognise and parse. This also logs wavfiles in $REGULUS/recorded_wavfiles.