Sphinx-4 Speech Recognition System
-------------------------------------------------------------------

Version:       1.0Beta+
Release Date:  xxx xxxx

Detailed Changes:
----------------

WDW - 05/20/05
    Incorporated contribution from Roger Toenz that fixes issues with
    fully qualified grammar names in JSGF.

PBL - 12/13/04
    Incorporated contribution from Gregg Liming that adds support for 
    dictionary addenda in the FastDictionary.

WDW - 11/09/04 
    Made WordPruningBreadthFirstSearchManager.recognize behavior
    consistent with the SimpleBreadthFirstSearchManager:  when
    their is no more data to decode, a null Result is returned.

    Made AccuracyTracker an abstract class; moved its best path
    logic to a new class: BestPathAccuracyTracker; created a 
    new class, BestConfidenceAccuracyTracker, that uses the
    ConfidenceScorer to determine the best hypothesis.

    Added a LatticeDemo to demonstrate how to obtain lattices
    from recognition results.


-------------------------------------------------------------------

Version:       1.0Beta
Release Date:  September 2004

-------------------------------------------------------------------

New Features:
-------------
    - Confidence scoring
    - Dynamic grammars
    - Narrow bandwidth acoustic model
    - Out-of-grammar utterance rejection
    - More demonstration programs
    - WSJ5K Language model


Improvements
------------
    - Better control over microphone selection
    - JSGF limitations removed
    - Improved performance for large, perplex JSGF grammars
    - Added Filler support for JSGF Grammars
    - Lots of bug fixes


Detailed Changes
----------------
PBL - 9/22/04 
    Improved performance of the DynamicFlatLinguist

    Upgraded the simple FlatLinguist to work properly with the
    WordPruningBreadthFirstSearchManager

    Fixed some problems with the PartitionActiveList that would allow
    incorrect tokens to be removed from the list during a replace.

WDW - 9/20/04
    Added selectMixer parameter and implementation to Microphone,
    which gives the ability to configure which Mixer the Microphone 
    class uses.  Helps with working around audio issue on the Mac.

PBL - 9/15/04
    Added scripts and instructions for building a WSJ5k language model
    from LDC data.

PPK - 9/13/04
    Completed confidence scores based on posterior probabilities.
    The MAPConfidenceScorer returns the confidence scores of best path
    and each word in the best path.

WDW - 9/10/04
    Added ECMAScript Action Tags support and demos.

PBL - 9/8/04
    Fixed bug 1003820 - Sphinx4 cannot handle non-ASCII characters in dictionaries
    Fixed bug 995124 - jsgf right hand recursion is now properly supported
    Fixed bug 981889 - Eliminated JSGF grammar limitations
    Addresed RFE  1012098 - add filler support to JSGF style grammars

PBL - 9/2/04
    Fixed bug 995123 - JSAPI allocate/deallocate issues

PPK - 8/25/04
    Fixed top level build.xml so that the various Java files for each
    acoustic model will not be recompiled over and over again.

PPK - 8/24/04
    StreamDataSource now throws an error if the 'bitsPerSample'
    is not a multiple of 8.

PPK - 8/24/04
    Fixed the lattice code by incorporating the acoustic and language
    scores into the lattice. Also modified the lattice creation code
    so that a lattice can be created from a full token graph, as well
    as the reduced token graph. Added code to test two lattices for
    equivalence.

PG  - 8/12/04
    Added sausage creation code in the result package.

PPK - 8/11/04
    Added the interface method WordSearchState.isWordStart() to
    indicate whether the WordSearchState is at the start or the end
    of a word.

PG  - 8/10/04
    Added posterior probability computation to the lattice.

WDW - 8/03/04 - 
    Added 8kHz models trained from the WSJ0 channel 1 training data:
    WSJ_8gau_13dCep_8kHz_31mel_200Hz_3500Hz.

PBL - 7/21/04 -
    Improved look of ZipCity, now with a USA map.

PBL - 7/20/04 -
    Fixed bug in JSGFGrammar parsing of NULL productions

PPK - 7/20/04 -
    Fixed bug in Microphone that causes an IOException when
    startRecording() is called immediately after stopRecording().
    The method stopRecording() will now wait until the audio line has
    actually stopped before it returns.

PPK - 7/19/04 -
    Added WavFile demo to show how to decode a .wav or .au audio file.

PBL - 6/30/04 - 
    Added the Sphinx-4 FAQ

PPK - 6/30/04 -
    Added stereo to mono audio converter to Microphone.

PBL - 6/21/04 - 
    Added the DynamicFlatLinguist. This linguist supports
    C&C style grammars (such as JSGF grammars). It differs from the simple
    FlatLinguist in that it can load much larger and more perplex grammars,
    loads them much more quickly and uses less memory. Currently this
    linguist runs slower than the simple FlatLinguist. Performance work is
    underway.

PPK - 6/21/04 - 
    Added CI Phone loop support for rejecting out-of-grammar utterances

PPK - 6/14/04 -
    Added ability to read different audio file formats to BatchRecognizer.

-------------------------------------------------------------------
Version:       0.1alpha
Release Date:  June 2004

-------------------------------------------------------------------

Initial release