sctk.htm 3.03 KB
<!-- $Id: sctk.htm,v 1.2 2004/08/30 14:54:03 jfiscus Exp $ -->
<HTML>
<HEAD>
<CENTER><TITLE>NIST Scoring Toolkit Version 2.0beta</TITLE></CENTER>
<B><BIG><CENTER> Welcome to the NIST Scoring Toolkit Version 2.0 Beta 1</CENTER></BIG></B>
</HEAD>
<BODY><p><hr>

The NIST Scoring Toolkit (SCTK) is a collection of software tools
designed to score benchmark test evaluations of Automatic Speech
Recognition (ASR) Systems.  The toolkit is currently used by NIST,
benchmark test participants, and reserchers worldwide to as a common
scoring engine.

<P> This is a beta version release of SCTK which now includes the
tranfilt software package.  The documentation has not been fully
integrated because of time limitations.  </P>


<P>
This version of SCTK contains several programs:
<P>
<B><A HREF="sclite.htm">sclite</A></B>

<UL>
<P>SCTK has at its core, sclite (Score-Lite), which is a flexible
Dynamic Programming alignment engine used to "align" errorful
hypothesized texts, such as output from an ASR system, to the correct
reference texts.  After alignment, sclite generates a veriety of
summary as well as detailed scoring reports.
<P>
This version of sclite comes bundled with the 
<A HREF="../src/slm_v2/doc/toolkit_documentation.html"> CMU-Cambridge Statistical Language
Modeling Toolkit v2.</A>  The toolkit is used to compute word-weights based on an N-gram language
model.  The directory 'src/slm_v2' contains the complete distribution and is automatically
compiled by the installation scripts.

</UL>

<B><A HREF="sc_stats.htm">sc_stats</A></B> <UL> <P>While sclite aligns
and scores a single system, sc_stats will compare system performance
between more than one system, so long as the systems under test have  been
ran on identical test data and using an identical test paradigm.  Inter-System
comparisons are made by running tests paired-comparison statistical
significance tests.  </UL>

<B><A HREF="rover.htm">rover</A></B> <UL> <P>Rover - Recognition
Output Voting Error Reduction, is a tool which combines an arbitrary
number for ASR system outputs into a composite Word Transition network
which is then searched an scored to retrieve the best scoring word
sequence.

<P>The program is documented in the paper <A HREF="rover/rover.htm">
A post-processing system to yield reduced word error rates: Recognizer
Output Voting Error Reduction (ROVER)</A> presented at the 1997 IEEE Workshop
on Automatic Speech Recognition and Understanding.  The paper is also
available in <A HREF="rover/rover.ps"> postscript.</a>

<B>hubscr</B> <UL> Wrapper script for scoring DARPA STT evaluations </UL>
<B>rfilter1</B> <UL> Transcription normalization filter engine </UL>
<B>csrfilt </B> <UL> Transcription normalization wrapper script </UL>
<B>chfilt </B> <UL> CallHome/Switchboard style transcript filter </UL>
<B>hamzaNorm </B> <UL> Arabic Hamza+alif transcript normalization filter </UL>
<B>acomp </B> <UL> Automatic compound word expansion filter </UL>
<B>def_art </B> <UL> Aribic definate article striping program </UL>
<B>utf_filt </B> <UL> UTF file format transcription filter </UL>


</BODY>
</HTML>