Didi Home Page

Didi (Dialog Displayer) is a freeware package for simple display, playback, and labeling of conversations and other speech signals.

This document describes in general terms what Didi is intended for. More information appears in the Users Manual and in the Limitations List. Didi Screendump

1. Purposes

The specific thing that Didi supports best is our own work, which involves looking for prosodic and voicing correlates of conversation phenomena, among other topics. More details are available at http://www.cs.utep.edu/nigel/. Since we study lots of things, we have kept the core code simple, to make it easier for each user to extend and customize (in C).

We have also used Didi for a labeling assignment in a course on Natural Language and Speech Processing.

2. Hardware and Software Requirements

Sun Sparcstation 5 or 10 etc, running SunOS 4.1, with X, /usr/demo/SOUND, or

PC with a soundcard or a sound chipset, running Linux, with X, and soundcard driver (standard or from 4Front Technologies)

3. Rival Packages

a Tcl/Tk -based tool for a wide range of tasks in speech research, freeware from KTH, available here
free from the LDC supports the transcription and annotation of audio recordings, available here
``a research, publication, and productivity tool for phoneticians'', freeware from the University of Amstermdam, available here
an industrial strength waveform analysis package, originally developed at Bell Labs, for Linux and Mac OS-X, downloadable from LDC here
the Computerized Speech Laboratory, sold by Kay Elemetrics, a PC-based hardware/software combination originally developed for a university phonetics laboratory.
the Speech Filing System, freeware from the University College London, again with a phonetics research orientation, available here
freeware from DGA, in France, "geared towards the transcription of long duration broadcast news recordings", available here Also with extensions to support meeting transcription at ICSI.
``acoustic analysis software'' with a ``carefully designed interface'', available here
``designed to facilitate the transcription of recorded speech'', for discourse analysis applications, shareware from the University of Santa Barbara, available here
for other tools see the LDC Annotation Page and Linux-Sound
Compared to these systems, Didi is relatively primitive, but it is good for its niche, namely the analysis of dialog phenomena.

4. Authors

Didi was written by Nigel Ward and incorporates contributions by Wataru Tsukahara and Yuichiro Fukuchi, all of the University of Tokyo. The Linux port was done by Wataru Tsukahara with follow-up by Rafael Escalante, Anais Rivera and Tasha Hollingsed of the University of Texas at El Paso.

Suggestions and bug fixes should be sent to nigelward@acm.org.


Up to Nigel's Home Page