New in version 1.4.0 (21 February 2011)
 * Fix conjunctions to work with current versions of link-grammar.
 * Fix anaphora printing for entities.
 * Fix build to look for OpenNLP-1.4 if OpenNLP-1.5 not found.

New in version 1.3.1 (15 January 2011)
 * Work around JWNL bug 3006600 for infinite loop -- 
   https://sourceforge.net/tracker/?func=detail&aid=3006600&group_id=33824&atid=409470
 * Add .n-f and .n-m linkgrammar gender tag recognition
 * Fix: anaphora resolution failure due to bad head-word identification.
 * Fix: missing frame concept error.
 * Fix: Add /usr/lib/jni to the library search path.
 * Add support for OpenNLP-1.5: launchpad bug 667949
   https://bugs.launchpad.net/relex/+bug/667949

New in version 1.3.0 (15 May 2010)
 * New date relationships for date tagging.
 * Fix: use of quantity in money expressions.
 * Downcase all flag names, so as to offer a more consistent usage.
 * Major restructuring of the entity maintainer interfaces 
   (this bumps version number)
 * Fix certain morphology handlers e.g.'d->would or 'd->had 
 * Misc frame rules fixes from Jim White, including a crash fix
 * Misc frame rules fixes from Fabricio Silva.
 * Clarify JWNL not found error message problems.

New in version 1.2.1 (3 November 2009)
 * Provide optional Penn-Treebank-compatible POS tagging.
 * Bug fixes for misc mis-identified POS tags
 * Bug fix for mis-identified advmod relation.
 * Improved word features, including comparative, superlative tagging, etc.
   IMPORTANT: This *requires* link-grammar-4.6.5 or newer!
 * Fix crash on opencog output module, on comparative sentences.
 * Fix: string substitution of "be" in frame rules (Fabricio)
 * Fix: correctly pass variable nodes to opencog (Fabricio)
 * Fix: printing of variables in the "simple view" (Fabricio)
 * Fix: misc pet-embodiment frame rule changes (Fabricio)
 * Move subscript-based feature tagging out of java code into algs file.
 * Remove usage of isPastTenseForm(), in favor of verb subscripts.
 * Fix: Fallback sentence splitter will use standard Java API.

New in version 1.2.0 (12 October 2009)
 * Fixes DocSpliter fallback to behave when it is misconfigured.
 * Fixes for anaphora resolution; head-word was missing or mis-identified.
 * Fix crash when the distributed parser (remote parser) returns null.
 * Don't print the phrase string in the OpenCog server, it messes the output!
 * Fix issuing of _amod when superlative modifier is involved.
 * Fix OpenCog Frame output to include truth values (Fabricio)
 * Change OpenCog Frame output to hold frame references (Fabricio)
 * Frames: add new concept vars and production rules. (Fabricio)
 * Use PrepositionalRelationNode in OpenCog prep relation output.
 * Tag verbs of questions as hypothetical.
 * Explicitly identify "filler subjects" aka "expletives"
 * Identify parataxis clauses.
 * Sort parses into ranked order.
 * Remove % from _%quantity relation name.
 * Remove garbage generated for certain prepositional sentences.
 * Remove garbage generated for certain questions, e.g. "Where is the ball?"
 * Fix cases where direct, indirect object reversed.
 * Fix handling of future tense in certain cases.
 * Fix faulty printing of certain numeric quantities.
 * Fix bug with multiple _predadj in one sentence.
 * Add a prototype Stanford-parser compatibility mode.

New in version 1.1.0 (2 July 2009)
 * Fix crash when generating scheme frames for opencog.
 * Distinct anaphora histories for distinct conversation streams. 
 * Fix extents calculation in lexical chunker code.
 * Requires new Java interfaces in link-grammar version 4.5.8

New in version 1.0.0 (17 June 2009)
 * Add "ant help" and "ant javadoc" build targets
 * Upgrade to opennlp-1.4.3, maxent-2.5.2
 * Move the old, deprecated parser API to parser-old.
 * Update for link-grammar version 4.5.7
 * Fix _amod, _predadj, _advmod order in the frame rules.
 * Fix lemma form of verb 's ("That's what the man said")
 * Fix handling of word subscripts, when subscripts are multi-letter.
 * Fix OpenCog output for frame elements.
 * Improve POS tagging coverage.
 * Strip the [?], [!] and [~] suffixes from link-grammar-guessed words.
 * Add refined sentence history support for anaphora resolution.
 * Export anaphoric reference output to OpenCog.

New in version 0.99.0 (May 2009)
 * Fix/reverse the usage of the head-word in certain relations.
   Please note that this is a MAJOR INCOMPATIBILITY from previous
	versions! Please note that the frame relations are BROKEN as
	a result of this change; they will be fixed later!

New in version 0.98.4 (March 2009)
 * Fix build break in optional OWL-format output.
 * Delete obsolete interface class RelexInfo.
 * Add getPOS() method to ParsedSentence.
 * Split relex-semantic-algs.txt into two files.
 * Correctly identify time-related adjectives.
 * Fix the way that the version number is identified during build.
 * Tag more pronouns as such.
 * Work around crash when input contains utf8-encoded Japanese/Chinese.
 * Add opencog-server.sh for generating OpenCog output.

New in version 0.98.3 (February 2009)
 * Fix bug, where the OpenNLP doc splitter was not being found.
 * Clarify installation instructions.
 * Tag certain date expressions as such.
 * Collect sentence statistics in the CFF output parser.
 * Fix parsing bug introduced in version 0.10.1

New in version 0.98.2 (January 2009)
 * Fix error in optional GATE/OpenNLP build.
 * Add support for automatically-propagated RelEx version number.

New in version 0.98.1: (Jaunary 2009)
 * Expose the new link-grammar "disjunct string" API
   (this requires link-grammar version 4.4.2).
 * Add comment delimiters to the frame rules file.
 * Fix, now supports jwnl-1.4rc2
 * Add initial OWL-format output.
 * Make install/use of GATE and OpenNLP optional.
 * Cleanup README to be clearer on how to use RelEx.

New in version 0.98.0:  (December 2008) 
 * Bump version number, in preparation for a 1.0 release.
 * New JSON interface to the link-grammar parser, for improved server
   performance (from Borislav Iordanov).

New in version 0.10.1:
 * remove cleanSentenceString(), replace with entity id's.
 * Avoid printing parens in the constituent string.
 * Add a pom.xml file for use with Maven.

New in version 0.8.4:
 * Convert DocSplitter to factory, so that the OpenNLP install is optional.
