The Meeting Recorder Tools

Tools for use with the Meeting Recorder corpus will appear here when they are ready to be released.

All tools are Copyright (C) 2003, 2004 International Computer Science Institute, and are provided AS IS. No support or warranty is offered.

Feel free to email requests, suggestions, bug reports, and patches to mrcontact@icsi.berkeley.edu.

mrt_tagConvert/remove tags from .mrt files
mrt2listConvert from mrt to a STM-like list format
soundmix  Creates an audio file by mixing several input audio files
XSLExtensible Stylesheet Language (XSL) converters for mrt files

Many people have requested additional tools for processing sphere/audio files. I suggest checking LDC's web page "Using LDC Data" at http://www.ldc.upenn.edu/Using/ (we have no association with the tools there, other than being users).

We are in the processing of rewriting the tool we used to transcribe the meetings ("channeltrans"). We will post the tool here when it is ready. We have also formed a Yahoo group for discussing the new tool: http://groups.yahoo.com/group/itranscribe/.

mrt_tag

This perl script converts .mrt files to and from the textual transcriber's conventions and embedded XML tags. It can also remove embedded tags completely. Only elements within the <Segment> and </Segment> tags in the .mrt file are effected.

Example:
Textual:  Hello '(6x) {GER} {QUAL whispered off-mike}
XML:  Hello <Foreign Language="de"> <Uncertain NumSyllables="6"> @@ </Uncertain> </Foreign> <Comment Description="whispered off-mike"/>
 Removed:  Hello @@

Requires XML::Parser (e.g. from CPAN).

Perl source as one perl file.

mrt2list

A perl script that converts a .mrt file to a simple stm-like list output. This program is not yet finished, and I only recommend it as a starting point for a full MRT => STM converter.

Example:
  > mrt2list Bmr003.mrt
  Bmr003 me022 chan2.sph 22.706 24.788 This is Eric on channel three, I believe.
  Bmr003 me011 chan0.sph 24.946 25.485 O_K.
  etc...

Requires XML::Parser (e.g. from CPAN).

Perl source as one perl file.

soundmix

This C program will mix a set of input audio files (for example, to mix the head-worn microphone signals for easier listening). It supports many different formats. You can either specify the gain of each channel, or the program can estimate the gains required to equalize the volumes (the auto-gain algorithm is fairly primitive).

Requires libsndfile, a cross-platform audio processing library by Erik de Castro Lopo available under the GNU Lesser General Public License from http://www.mega-nerd.com/libsndfile.

C source as one C file. As the dependencies are really minimal, autoconf and Makefiles are not provided.

XSL Style Sheets for Transcript Files

Maël Guillemot at IDIAP has written some xsl tools for converting transcripts from XML to HTML. The tools would also be a great starting point for other types of conversion. The tools can be found at http://mmm.idiap.ch/documentation/XSLtools/.