[sfrechtling]

Yes, some NLP POS (part of speech) taggers do use context. I'm not entirely sure if they work in the way that you describe. I believe that they use the grammar of the sentence to derive what the tags over the sentence are. That is, they can tag the sentence on what they know and then iterate over the sentence until everything can fit into the grammar (or corpus) that it was trained on.

If you want more background - the nltk book, in particular chapter five is a good place to start: http://www.nltk.org/book/ch05.html

What library did you use? NLTK on Python 2.7 gave the following:

[('We', 'PRP'), ('will', 'MD'), ('project', 'VB'), ('our', 'PRP$'), ('earnings', 'NNS')]

[iLoch]

Interesting I'll have to take a look at this Python lib. I was trying out various Node.JS libs.