2002-11-08  Evandro Gouvea (egouvea@cs.cmu.edu) at Carnegie Mellon University.
	- Converted Makefiles to use GNU's autoconf, automake, libtool etc.
	- Added acoustic model, language model, dictionary, etc to use as
          an example system in a release.
        - Release of the Sphinx-3 package.

2001-01-01   Rita Singh (rsingh@cs.cmu.edu) at Carnegie Mellon University
	- Introduced live-mode interface
	  
2000-02-29   M. K. Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
	- Changed compile-time selection of 3 or 5-state HMM (with two 
	  executables) into a single binary that does a run-time selection.
	- Makefile changes to go with the above

2000-02-12   M. K. Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
	- Consolidated all code under a single directory.
	- Removed SubVQ-PhoneProb (svqpp) hack for prediction of active phones.

1999-12-17   M. K. Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
	- Added src/libcontmgau/gausubvq.c for sub-vector clustering of 
	  acoustic models, and related changes to libcontmgau.

1999-12-13   M. K. Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
	- Added ability to process a single utterance over and over, 
	  waiting for file timestamp (modification time) to change 
	  between iterations.
	- Added new output format for lattice files, for smaller size.
	- Added SubVQ-PhoneProb module for prediction of active phones.
	- Added special handling of 3-state, no-skip topology HMMs.
	- Fixed bug in hmm-evaluation of non-emitting exit state scores.

1999-11-23   M. K. Ravishankar (rkm@cs.cmu.edu) at Carnegie Mellon University.
	- Initial release.
	- Lextree decoder using continuous density HMMs.
	- Batch mode only.
	- Cepstra input only.
