Address extraction code is impossible to work on

[slinkp]

The regular expression that ebdata.nlp.addresses uses to find addresses is actually a 100-line regex into which a 130-line regex is inserted 11 times. The final regex is over 1500 lines long.

It is almost impossible to debug, fix, or extend this regex.
We need to re-think the address extraction approach completely.
Investigating whether there is existing natural-language work we can leverage.

[slinkp]

Ticket imported from Trac:
http://developer.openblockproject.org/ticket/128
Reported by: slinkp