Numbers v1.3
FINAL RELEASE - Release Version 1.3 (23 August 2002)
There is 12618 speakers in 23902 speech files we added time aligned phonetic labels.
Release Version 1.2 (3 June 2002)
We add more files (now 12618 speakers in 28626 speech files) and there were made some changes in directory structure.
Release Version 1.1 (18 August 2000)
Several changes have been implemented that differentiate this version of the Numbers Corpus from version 1.0. These changes include the following:
- Speech files in the /speech directory have been converted from NIST format to RIFF format.
- Individual transcription files have been extracted from the trans.txt file. These individual files have been placed in /trans directory which has a structure that exactly parallels that of the /speech directory.
- The documentation has been updated so that it accurately reflects the corpus contents.
These and other changes have been made to make the corpus more useful to the end-user.
Version 1.0cd (June 1998)
This corpus was originally collected and distributed on 8mm tape for use on UNIX-based workstations. For this release, we have converted the corpus to a more platform-independent file structure.
- Converted file and directory names to 8.3 naming format.
- Verified mulaw format for all audio files.
- Rewrote audio file headers with file name and corpus information.
- Updated documentation to reflect changes.
- Burned to ISO 9660 CD-ROM.
Version 1.0
- First complete release.
