Changelog

version 3.20 / 2019.11.30
Fixed bug in XMLtext.cpp that caused words containing character references not
to be decoded properly. Fixed bug in sst.cpp that caused a segmebtation error
with some inputs.

version 3.12 / 2019.11.19
Changed (CONST)VOIDP to (const) char* in some cases.
Removed a test from an alternation that duplicated another test in the same
alternation, but that also imposed an extra condition, having no effect.

version 3.11 / 2019.07.09
Added decoding of numeric character references and html character entity 
references.

version 3.10 / 2017.06.07
Fixed: core dump when input started with empty line(s)
Fixed: option -x- (no xoptions file for extra options)

version 3.9 - 2017.03.24
Fixed bug in fst.cpp: write output before freeing contextualRules.

version 3.8 - 2017.02.13 
Option -x- suppresses the use of the 'xoptions' file. (Use -x- together with
-p and -n options.)

version 3.7 - 2016.11.09 
Enlarged scope of tag_hash.

2016.05.17

Start of changelog.
There have been several versions of taggerXML before this changelog was
started.

taggerXML is based on Eric Brill's part of speech tagger. The original C-code
is changed to C++ code and made platform-agnostic. (The original code was for
UNIX platforms only.)
In this version a number of improvements have been made. Janez Brank suggested
a number of improvements to get rid of memory leaks. These have all been
implemented.
The program could crash if the last line of text input (one sentence-per-line)
didn't end with whitespace. This is solved.
The program was not stable if #define STREAM 1 (in defines.h). This is solved.
A lot of commented code and useless comments were removed. There is still a
lot left.