NEWS

PocketSphinx 5.0.3
^^^^^^^^^^^^^^^^^^

The NEWS file has not been updated for some time.  See the GitHub
releases page at https://github.com/cmusphinx/pocketsphinx/releases
instead.

Pocketsphinx 0.8
^^^^^^^^^^^^^^^^

Thanks: Tanel Alumae, Douglas Bagnall, Halle Winkler, Peter Grasch, Vyacheslav Klimkov
Erik Andresen, Nicola Murino, Melmahdy, Scott Silliman, Riccardo Magliocchetti, Marc Legendre,
Kho-And-Mica, Zheng6822, Pankaj Pailwar, Evandro Gouvea

Improvements:
 * API is better exposed in Python, SWIG and gstreamer
 * New API for exact grammar match
 * Configurable type for the frame counter, allows utterances way longer than 5 minutes

Fixes:
 * Memory leaks, refcounting and other memory-related issues
 * Use proper word boundary senones for the words added on the fly
 * Accurate FSG lextree construction

Pocketsphinx 0.7
^^^^^^^^^^^^^^^^

Thanks: David Huggins-Daines, Chen Tao, creative64, Edwin Miguel, Bhiksha Raj

Improvements:

 * Nice performance improvements
 * Use arc iterators to build FSG lextrees, speeds this up immensely
 * Filler word propagation is responsible for 5% speedup
 * Continuous decoder can decode files, not from the microphone
 * Add senone score dumping and reloading capability
 * Functional alignment pass
 * Symbian support 
 * Add the Android makefile and other JNI stuff
 * Case sensitive names, including phone names
 * Better error messages
 * Report timing for individual search passes

Bug fixes:

 * Accuracy regression fixes
 * Fixes the bug with reading transition matrix on big-endian platforms
 * Fixes very important regression with NULL transitions in fsg search
 * Proper acoustic score scaling during posterior calculation.
 
And many, many, many more interesting things!

Pocketsphinx pre
^^^^^^^^^^^^^^^^

2000-01-27 CMU Sphinx is going Open Source!