Personal tools
You are here: Home → Documentation and release notes
 

SPRAAK documentation

When SPRAAK is installed on your local machine, part of the installation process consists of extracting the documentation from the source files. After this process has finished, the reference manual can be found in the doc/html directory. Open the index.html file in this directory with your favorite browser to consult the manual of your installed version.

The documentation of the most recent release of SPRAAK can be found here.

An explanation of the newest features in version 1.0 can be found here.

IMPORTANT: at registration time licensees received a download key in the form of a URL.
If this URL starts with

http://fettercairn.esat.kuleuven.be:8080/Spraak/spr_register/download?key=

 then you have to replace this prefix with

http://www.spraak.org/spr_register/download?key=

 Users who received a license recently don't have this problem.

 

Release notes for v1.0:

  • improved documentation
  • tutorials
  • new examples using the TIMIT database
  • the default file format is now the SPRAAK (.spr) format
  • added code for exemplar based recognition (DTW based rescoring)
  • added code for neural networks (MLP)
  • added code for FLaVoR decoding (phone lattice to word recognition)
  • new Matlab read/write routines
  • spr_version.py makes it easy to add SPRAAK to a login-script or to switch between version
  • extra global options to all programs to control the nice level, to make subdirectories on file creation, ...
  • spr_lm_arpabo now automatically fills in the gaps in a given arpa-style N-gram
  • new/improved preprocessing modules: [histeq], [ordstat]
  • high-level support to put phone/state information in word lattices
  • the legacy arcd files are no longer needed, can now be replaced by cd files
  • spr_fv_gauss now also outputs fMLLR statistics
  • various small improvements and corrections
Document Actions