Changes

Reminder: 'state_city ' is the states extracted from 'city'; 'state_addr1 ' is the states extracted from 'addrline1'; 'state_addr2 ' is the states extracted from 'addrline2'.

The 'state_city', 'state_addr1 ' and 'state_addr2 ' are consistent.

One reason for the inconsistency between 'state ' and ('state_city', 'state_addr1 ' and 'state_addr2') is the co-existence of abbreviation and full form. So they are the same actually. Otherwise, 'state_city', 'state_addr1 ' and 'state_addr2 ' beat feature 'state ' because they are detailed.

For records of which 'addrline1', 'addrline2 ' and 'city ' don't contain state info, just take feature 'state ' as the 'state_cleaned'.