Soothsayer 0.6.2 released

Soothsayer 0.6.2 is now available for download.

Soothsayer 0.6.2 comes with a number of new features. Most notably, a new statistical predictive plugin, based on recency promotion, is available. The new recency plugin generates predictions by assigning exponentially decaying probability values to previously encountered word tokens, thereby promoting context recency.

Soothsayer 0.6.2 also ships a brand new simple GUI demonstration program, prompter. Prompter is a soothsayer-enabled text editor. Prompter displays predictions generated by soothsayer through a pop-up autocompletion list. Prompter also provides an autopunctuation feature that saves key pressing by intelligently handling punctuation and whitespace. Prompter is a Python application (wxPython) and uses soothsayer’s python binding.

Soothsayer 0.6.2 adds native Windows support by supporting the MinGW/MSYS platform. It is now possible to build soothsayer in native Win32 mode. Detailed instructions to build soothsayer on MinGW/MSYS are included in the doc/ directory.

Soothsayer 0.6.2 includes enhancements to the build system, a restructured soothsayer exception hierarchy, additional range checking in core classes, and improved logging subsystem.

Soothsayer 0.6.2 also includes a number of bug fixes. See ChangeLog for more details.

Soothsayer 0.6.2 is a beta release. This is a source release only. No precompiled packages or installers are provided.

Users wishing to try out Soothsayer will need to follow the (easy) steps required to build soothsayer on their machine, as detailed in the README file. Please note that SQLite is required to build soothsayer. CPPUnit is optional, but be aware that no unit tests will be built nor run when running `make check’, unless CPPUnit is installed.

Soothsayer 0.6.2 has been built and tested on various Linux platforms (including 32-bit and 64-bit architectures), Solaris 10, Windows XP/Cygwin, and Windows XP/MinGW/MSYS. If you encounter any issues while building or running soothsayer, please report it to the author.

As always, there is still a lot more work to be done. Currently, the installed soothsayer system is trained on a very small training corpus. Predictive performance can be greatly increased by using a larger training corpus. Users can easily generate statistical predictive resources using the text2ngram tool on a custom training text corpus.
New predictive plugins are also in the works, which will take advantage of the multiple predictive source architecture.
Please refer to our TODO list for details on what needs to be done.