Parse addresses with no address number and two words as street name

[bbharathrao]

Training for the address with no address number and two words as a part of street name.
Example 1:

The Way it is parsing currently

address1 = usaddress.tag("Mid Island Street HICKSVILLE NY 11801")
pprint(address1)
(OrderedDict([('AddressNumber', 'Mid'),
('StreetName', 'Island'),
('StreetNamePostType', 'Street'),
('PlaceName', 'Hicksville'),
('StateName', 'NY'),
('ZipCode', '11801')]),
'Street Address')

The Way it needs to be Parsed

address1 = usaddress.tag("Mid Island Street HICKSVILLE NY 11801")
pprint(address1)
(OrderedDict([('StreetName', 'Mid Island'),
('StreetNamePostType', 'Street'),
('PlaceName', 'Hicksville'),
('StateName', 'NY'),
('ZipCode', '11801')]),
'Street Address')

Example 2:
The Way it is parsing currently

address1 = usaddress.tag("New Park Rd West Hartford CT 16110")
pprint(address1)
(OrderedDict([('AddressNumber', 'New'),
('StreetName', 'Park'),
('StreetNamePostType', 'Rd'),
('PlaceName', 'West Hartford'),
('StateName', 'CT'),
('ZipCode', '16110')]),
'Street Address')

The Way it needs to be Parsed

address1 = usaddress.tag("New Park Rd West Hartford CT 16110")
pprint(address1)
(OrderedDict([('StreetName', New Park'),
('StreetNamePostType', 'Rd'),
('PlaceName', 'West Hartford'),
('StateName', 'CT'),
('ZipCode', '16110')]),
'Street Address')

Training xml located at:
usaddress/training/double_street_name.xml

Testing xml located at:
usaddress/measure_performance/test_data/test_double_street_name.xml

[xmedr]

@bbharathrao I can definitely see that there are inconsistencies with how usaddress parses multi word street names like you describe. However this pr seems to have more training data for hcr boxes like #259.

If you happen to have more examples of multi word street name addresses that are being parsed incorrectly, then I can add them to the dataset! I'll close this pr for now.