Reimplement DKPro Core wrappers for Stanford CoreNLP based on CoreNLP annotators (1.12.0)

[reckart]

This is a follow up to #748 for new features that are implemented after 1.9.0.

Upgrading CoreNLP in DKPro Core at times was a bit painful in the past as we duplicated quite a bit of the functionality provided by the CoreNLP annotator classes by using pre-CoreNLP low-level classes directly. I hope reimplementing DKPro Core wrappers for Stanford CoreNLP based on CoreNLP annotators will facilitates upgrades in the future.

Another benefit of these new wrappers is, that they can use the multi-threading support built into CoreNLP.

Organizational

• [X] (1.9.0) create new corenlp module
• create new build.xml with updated model IDs (corenlp instead of stanfordnlp) and use them - should be based off per-language model JARs that are to be provided with CoreNLP 3.6.0 and following - Sticking with the models for stanfordnlp model for the moment.

Components

• [X] (1.9.0) TokenizerAnnotator + WordToSentenceAnnotator -> CoreNlpSegmenter
• [X] (1.9.0) POSTaggerAnnotator -> CoreNlpPosTagger
• [X] (1.9.0) MorphaAnnotator -> CoreNlpLemmatizer
• [X] (1.9.0) NERClassifierCombiner -> CoreNlpNamedEntityRecognizer
• [X] (1.9.0) ParserAnnotator -> CoreNlpParser
• [X] (1.9.0) DependencyParseAnnotator -> CoreNlpDependencyParser
• [ ] DeterministicCorefAnnotator -> CoreNlpCoreferenceResolver (in progress, need to rename)
• [ ] CorefAnnotator
• [ ] EntityMentionsAnnotator - could use NamedEntity annotations
• [ ] CoreNlpSentimentAnnotator - a bit tricky because it requires binarized trees and we cannot create these locally in the sentiment annotator: requires headfinder and other things available only in the parser
• [ ] CleanXmlAnnotator - not sure if we need that, might consider it as a normalizer
• [ ] RegexNERAnnotator - not sure if we need that
• [ ] TrueCaseAnnotator - not sure if we need that, might consider it as a normalizer
• [ ] RelationExtractorAnnotator - we have no types for this yet
• [ ] QuoteAnnotator - we have no types for this yet
• [ ] NaturalLogicAnnotator - we have no types for this yet (models only available when CoreNLP 3.6.0 is on Maven Central)
• [ ] OpenIE - we have no types for this yet (models only available when CoreNLP 3.6.0 is on Maven Central)

Upstream bugs

• [ ] Factored models broken in 3.6.0 (https://github.com/stanfordnlp/CoreNLP/issues/195)

[reckart]

Everything that would be converted for now has been converted. The old StanfordNLP module has been removed and we only have the CoreNLP module now.