cubrady / sctk

Speech Recognition Scoring Toolkit

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

File:  README
Date:  November 10th, 2009
SCTK Version: 2.4.0

asclite compression algorithm update.

This directory contains SCTK, the NIST Scoring Toolkit.  SCTK contains the 
source code, documentation and example data for the following programs:

	sclite   V2.3  - "Score Lite",
	sc_stats V1.3  - sclite's Statistical Significance Comparison tool
	rover    V0.1  - Recognition Output Voting Error Reduction
	asclite  V1.10 - Multidimensional alignment replacement for sclite

The distribution was initially written in ANSI C on a SUN Workstation.  To date,
the code has been successfully compiled and tested on the following platforms:

					      SCTK
Systems                                     Version 
---------------------------------------------------------------------     
SUN SunOS 4.1.4 using gcc                     
SUN Solaris SunOS 5.4 using cc                
SUN Solaris SunOS 5.5 using gcc               1.0 - 1.2
SGI IRIX 5.3 using cc                         
SGI IRIX 4.0.3H using gcc                     
HP 735 using gcc                              
HP using "cc"                                 1.0 - 1.2
DEC Alpha running DEC OSF/1 V2.0 "cc -std1"   
DEC Alpha running DEC OSF/1 V2.0 "gcc"        1.1
IBM Powerstation running AIX V3.2 using cc    1.0
IBM Powerstation running AIX V3.2 using gcc*  1.0
Linux Version 2.0.33                          1.0 - 1.2
Cygwin                                        1.3 - 2.0B2
Linux Version 2.2.17                          1.3 - 2.0B2
Linux Version 2.4.20-20-9 **                  1.3 - 2.1.7
Linux Version 2.6.19.fc5                      2.2.0 - 2.2.4
Linux Version 2.6.9-5.EL (gcc 3.4.6-3)        2.2.0 - 2.2.4
Darwin Version 7.4.0                          1.3
Darwin Version 8.5.0 (gcc v4.0.1 build 5367)  2.1.2 - 2.3-rc1 
Darwin Version 8.11.0 (gcc v4.0.1 build 5367) 2.3-rc2 - 2.4.0
Darwin Version 9.6.0 (gcc v4.0.1 build 5490)  2.3.6 - 2.4.0
RedHat Fedora 9 (gcc v4.3.0-8)                2.3.9 - 2.4.0

 * compiled with warning messages, but successful self test
** compiled and ran, but test6, which uses GNU's diff for alignment fails.  The 
   use of DIFF for alignment is rarely used and the failure can be safely
   ignored.

Read install.txt to compile and install the package.  UNIX-style manual pages 
and 'HTML' formatted manuals are provided in the 'doc' directory.

			Bundled Software:

This version of SCTK includes the CMU-Cambridge Statistical Language Modeling 
Toolkit v2.  Included in STCK is the complete source code tree and 
documentation.  The toolkit has been included so that SCLITE can use a language 
model to compute conditional word weights for word-weighted scoring.


			Optional Software:

To use the interactive installation for sclite, edit the file 
src/sclite/config.sh at line 36, set the user_interaction=1 it will ask you 
during the configuration (make config) if you want to use GNU diff instead of 
sclite and CMU-Cambridge SLM for sclite.  Those choices are not recommended and 
not used anymore.

Sclite has the ability to use GNU's "diff" utility as an alternate string 
alignment method.  The installer has the choice whether or not to include this 
ability in the compiled version.  

If the use of GNU's "diff" is desired, you will need GNU's "diff" utility from 
their "diffutils" distribution version 2.7.  If you cannot locate this 
distribution, it is available via anonymous ftp from the jaguar.ncsl.nist.gov 
server.  The compressed tar file 'diffutils-2.7.tar.Z' is located in the 'pub' 
directory.  When you install diff, make sure the executable is installed as: 
"gdiff", "gnudiff" or "diffgnu".  The 'config.sh' script expects it to be 
called one of these.

Special Note to SGI owners: When installing GNU's "diff" make sure to call it 
"gnudiff" or "diffgnu", if you use "gdiff" the executable name may be 
overshadowed by a pre-existing executable.

utf_filt is using SGML external software and modules like SGMLS perl module and 
nsgmls program which have be updated and continued: OpenSP 
(http://sourceforge.net/project/showfiles.php?group_id=2115&package_id=2869).  
The version v1.5.2 of OpenSP have been tested and working with the current 
version of utf_filt.


			64 bits Compilation

With big alignments, sctk needs to be compiled in 64 bits.
By default, the C/C++ software are compiled in 32 bits with the options (-Os) 
but can be compiled in 64 bits, -m64 is added to the CFLAGS variable in:
src/asclite/core/makefile
src/asclite/test/makefile
src/rfilter1/makefile
src/sclite/makefile
Example of CFLAGS:
For OSX 10.4+: -fast -m64 -arch x86_64 -arch ppc64

About

Speech Recognition Scoring Toolkit


Languages

Language:C 41.0%Language:C++ 19.7%Language:Shell 17.2%Language:Perl 12.3%Language:HTML 5.8%Language:JavaScript 1.7%Language:Makefile 1.0%Language:Prolog 0.8%Language:Perl 6 0.2%Language:Gnuplot 0.2%Language:Erlang 0.1%Language:Lex 0.0%