How to deal with parentheses?

[simongray]

I honestly have no idea, will need to think about it.

[simongray]

Here's an example:

// TODO: double subjects, confusion caused by parentheses)
String example = "I sure was (I come from Copenhagen, Denmark).";

creating this output:

I sure was (I come from Copenhagen, Denmark).
    |_ statement: {Statement: "I sure was -LRB- I come from Copenhagen, Denmark -RRB-", components: 4}
        |_ component: {Subject: "I"}
        |_ component: {IndirectObject: "from Copenhagen, Denmark"}
        |_ component: {Verb: "sure"}
        |_ component: {Subject: "I"}

The dependency graph is not very helpful in this case, unfortunately:

[sure/RB
  nsubj>I/PRP
  dep>[was/VBD
       dep>[come/VBP
            punct>-LRB-/-LRB-
            nsubj>I/PRP
            nmod:from>[Copenhagen/NNP case>from/IN punct>,/, appos>Denmark/NNP]
            punct>-RRB-/-RRB-]]
  punct>./.]

[simongray]

It seems like I could create a custom annotator after the tokenisation stage which

• removes noise such as emoticons entirely (tokens with non-alphanumeric content that is not treated as punctuation, e.g. ":-)", probably treating emoticons as a sentence divider
• removes - or in the future extracts - content between left and right parentheses
• if extracting, creates new annotations from extracted content

[simongray]

Perhaps the above is too complex for the time I have left and simply preprocessing to remove parentheses and emoticons is the better solution.