353
edits
Changes
Jump to navigation
Jump to search
Patent Assignment Data Restructure (view source)
Revision as of 15:52, 20 March 2017
, 15:52, 20 March 2017→Restructure Address Information
::*Five features (addrline1, addrline2, city, country, postcode) in the table contains address information.
::*Features addrline1, addrline2 and city are not cleaned. They have city, country and postcode information.
::*The object is to extract city, country and postcode information from the three features above.::*By now, we only focus on and clean cleaning American patents.
:2. Postcode Extraction:
::U.S. post code follows the pattern [five digits] or [five digits - four digits]. In this way, U.S. patents can be extracted by searching for post code following these patterns using regular expression. Some other countries also use [five digits] for post code, so only post codes following [five digits - four digits] are extracted.
::SQL code are in:
E:/McNair/Projects/PatentAddress/RxPostcode.sql
::The extracted records are stored in table ptoassigneend_missus. Applied similar methods to filter out patent records from Japan. The post code in Japan follows pattern [three digits- four digits]. The post code extracted is quite accurate for U.S., and so is the country information.
The problem is that the city information extracted is not quite good. It messes up with street names. One approach to increase the accuracy is to list all the possible cities in each country, and then match the address columns to these cities, which is time consuming.
