Performance and parsing improvements

Tagging a release following the work done @cnnorris on the original parsing method.

Minor updates

This release is to capture the state of PETRARCH before major revisions occur.

0.2.0

This is our stable, seemingly bug-free version of PETRARCH. We've run this (most of) this version for months now with no errors popping up. This is the version generating the most recent data seen on http://phoenixdata.org/.

What works?

Basically if you use this the way we do in production (in conjunction with the phoenix_pipeline) then everything should be great. If you try to use it as a standalone program you might hit some weirdness. The program should work as a standalone if you pass in parse trees from CoreNLP. The built-in integration with CoreNLP, i.e., passing in unparsed sentences and having these parsed as part of the PETRARCH coding process, doesn't exist right now. That feature also isn't of the highest priority for us currently. If you desperately need this functionality send us an email and we'll see what can be done.

Why not v1?

Because we haven't built in all of the ideal features yet. The standalone program needs a fair amount of work. Tests need to be written, etc., etc.,

0.1.0

Moving closer to 1.0.0. Many of the major bugs have been addressed, but some additional coding features need to be implemented.

Beta

Moving up to beta. Fix some issues such as trapping (NEC codes.

Alpha release

Alpha release associated with the alpha version of the Phoenix dataset. PETRARCH works pretty much the way it will moving into the future. There are a few internal coding features left to implement and some refactoring to be done, but it's most of the way there.