Top of Page
home | introduction | research | people | facilities | events & news | visitor info | contact us | search

The Spoken Childes Tool

Corpora with Speech Signal 

For psychologists as well as linguists whose work is closely related to corpora and who need to have quick and seamless access to the speech signal a set of new tools is available. Necessary preconditions of this feature are: 

  • that the speech information has been digitized beforehand 
  • that it is accessible on-line 
  • that all references between the text and the speech waveform have been correctly set. 
Currently great efforts are being made at various places, including the MPI für Psycholinguistiek, to do exactly this for many corpora included in Childes and in ESF. In the widely accepted CHAT-format this means that a speech tier must be created for every utterance. This speech tier provides information about the corresponding speech file and the begin- and end-points of the speech segmdnt for each utterance. 

CLAN

The CLAN editor programmed at CMU (Pittsburg) allows a MAC and Windows user to see and listen to the speech signal which corresponds to the piece of code identified by the position of the text cursor. This editor offers some simple sound related functions to operate on the speech signal such as listen, go forward, and go back. The CLAN editor is the successor of the CED editor and CLAN tools that have now been integrated in one programme.

CED - XWAVES Combination

A "Spoken Childes Environment" has been developed at the MPI which combines the XWAVES speech software and the CED CHILDES EDitor. This CED-Waves extension can be used to work with CED on a corpus and start the XWAVES speech software to be able to see the corresponding segment in the speech wave. Of course, the XWAVES environment gives you all its standard functionality, i.e. it gives the researcher all possible analytical tools at hand which are standard in a professional speech laboratory. Again the text and speech wave cursor are tightly linked to allow smooth operations. 

However, this XWAVES extension works only on UNIX systems. 

Platform Independent Spoken Childes Environment (MED)

A new version of what is called "Spoken Childes Editor" or MED has been implemented  in a platform-independent manner. It is implemented with the help of the TCL/TK script language and also incorporates some easily portable C-code. It offers a new type of WYSIWYG editor which also offers menu selections. It also offers a seperate waveform widget to display the speech waveform which has a number of functions such as scrolling left or right, listening to a segment of the speech waveform, zooming in or out, showing the F0 curve of a segment, and transcribing/labeling speech with orthographic or phonetic symbols. 

The cursor control is such that scrolling in the corpus means also scrolling in the speech waveform and vice versa, i.e. there is a tight link between text and speech wave cursor. 

In addition, the editor supports all major modes and major functions of the "old" CED editor. 

This tool is available for W95, Win NTand UNIX. A major difficulty for achieving full-platform independence was and still is to get a unified IPA character set and compatible fonts for the different platforms. To give Windows users access to speech processing software a link with the Praat programme was implemented. Unix users have the choice between Praat and XWaves.

The following picture gives an impression of the new "Spoken Childes Editor" with the waveform display window. In the code window the actual utterance is highlighted in yellow. It contains the so-called speech tier which is indicated by "%snd" following a loose version of the CHAT syntax. This tier contains all information necessary to activate the corresponding speech wave window. The red lines in the speech window show the bounderaies of the utterance. 

Last updated: February 14, 2000 17:23
End of Page