It still works fine inside a debian-docker container using XQuartz on
the host system.
Windows will be tested soon.

- Using 2to3 and some manual labor
   - Especially comments can be overlooked
- Updating lex.py and yacc.py
- Tested the changes with arabic.ccg
- The editor buttons are broken with this version

Additionally removed some outdated imports and comments.
One "arg == None" was changed to "arg is None".

Update NERApp to be compatible with current Stanford Core NLP

Got Ant build errors due to 'includeantruntime'

added links to @dmhowcroft's introductory posts

Also did some formatting cleanup

Not doing anything clever or changing behavior, just added in 'false' for the new arg.

See https://github.com/stanfordnlp/CoreNLP/blob/master/src/edu/stanford/nlp/ie/NERClassifierCombiner.java#L103 for the method definition in question, and the same method def in Stanford Core NLP v. 1.3.4 for comparison.

From StackOverflow: https://stackoverflow.com/questions/5103384/ant-warning-includeantruntime-was-not-set/5103432#5103432

Duplicate subexpression in Slash.java

`_ability == ACTIVE` is repeated.

Added option to allow categories with missing index nominals on their target cats to combine during realization, as an exception to the index filter, so that type-raised cats involving sibling dependencies can combine prior to combining with the head cat.

Remove warnings (unclosed readers/writers and unsafe varargs)

Changed handling of ccgbank/extract to ignore all files (with exceptions) rather than the whole directory itself.

Added note re lack of guarantee of lex lookup returning the right arg roles if word-cat pair is observed.