Add sentiment attributes to XML output

[relwell]

The following changes are included:

• Update XmlOutputter for sentiment: this allows us to use sentiment in the default pipeline
  • Add a "sentiment" attribute to each sentence node for the
  • Add an "averageSentiment" attribute to the sentences node
• Fix static method call in MorphaAnnotator -- just a quick fix from an IDE complaint
• Add sentiment to the default Stanford CoreNLP pipeline classpath properties

[AngledLuffa]

Thanks for the suggestion! It is not often people make pull requests, so there should be some sort of prize available.

I'm not sure the average overall sentiment in a document is that relevant if calculated in such a manner. Mostly I'm skeptical that the final result is meaningful, since it just says whether positive or negative sentences are more common.

Also, since the domain of the model is relatively limited compared to the other tools (movie reviews vs all of WSJ), I'm not comfortable turning it on by default.

The other changes have been incorporated, though.