Added new steps

[aalexandrasilva]

• Restructured the project. Mostly organized the files to the packages by component. For example, all claim checkers should be inside the package claim_worthiness_check.
• Added coreference resolution as a REST service with spring.
• Integrated coreference resolution with our pipeline. It is now between the translation and the claim check step.
• Added multiple configurations that we attempted to the first step of WISE such as scaling and over-sampling from minority classes.
• Added the RNN training script and integrated the trained model for the final prediction.
• Changed the output format to a better-formatted JSON.
• Added the result of the indicator check to our output.
• We also now create the database if it isn't there when starting the service, instead of running 2 separate scripts.

In this file, there's an example output of the service for /rawstatus and this file

there is the output for /status.

[MichaelRoeder]

A lot of nice changes. However, I couldn't resist to add some comments 😉

[MichaelRoeder]

We do not really need this Java service, or? We can simply use a dedicated server of the CoreNLP project (https://stanfordnlp.github.io/CoreNLP/corenlp-server.html#dedicated-server).

[aalexandrasilva]

As far as I remember from what @umairq mentioned, we needed it because CoreNLP only finds the coreference chains and doesn't replace them. This is the coreference resolution that was present in the FactCheckUtilityTools repository.
Maybe we can also use their dedicated server and do the replacement in the python side. This is only worth it if we insist on keeping Stanford's coreference. We also have the spacy implementation now in https://github.com/dice-group/NEBULA/blob/wise/org/diceresearch/nebula/coref_resolution/spacy_coref.py.
But maybe @umairq can shed some light there.