How does the Stanford Parser handle unseen words?

Question

I'm curious to know more about the heuristics/techniques used by the Stanford parser for handling unseen words.

For example, given input: aa aaa aaaa, "bb bbb bbbb"?, the parser produces: (ROOT (SQ (VP (VB aa) (NP (NP (NN aaa) (NN aaaa)) (, ,) () (NP (NN bb) (NN bbb) (NN bbbb)) ('' ''))) (. ?))). As another example, given: aaa aaaa, the parser output is: (ROOT (NP (NN aaa) (NN aaaa))). How does the parser parse these sentences?

I've looked on the Stanford Parser FAQs page as well as the parser mailing lists, but couldn't find any discussions on this topic.

Thanks in advance for any insights.