Inventor U-turns & by-hand assignee consolidation

Code and data for detecting "u-turns" among inventors. If two inventors have been "lumped" into a single inventor ID, or if a single inventor patents concurrently with multiple assignees, it may appear that the inventor is moving repeatedly when in fact something is going on. Uturns.py detects these "u-turns" in the patent data and labels them so that researchers can either exclude them from inventor mobility tables or check them manually. The input file Order.inventor.geo.assignee.combo.disambig.txt.7z is a compressed file of patent data with the pipe-delimited format patent #|inventor ID|assigneeid|application date| internal ID | grant date | inventor name | last name | first name | city/state | country | state | city | inventor # on patent | inventor ID | county | fips | fips state | fips county | latitude | longitude | zip code | msa. The output file is pipe-delimited with patent # | inventor ID | internal ID | application date | grant date | inventor name | last name | uturn identifier. Note that records are only output by Uturns.py for inventors with u-turns. 

Also, patent assignees are notoriously difficult to consolidate due to slight variations in naming. The sheer number of assignee variations make it impossible to check all by hand. However, we use the aforementioned u-turns to identify possibly-missed consolidations, i.e. if an inventor u-turns between two companies with very similar names. We use soundex to identify very similar names within u-turns and then check all of these by hand. The resulting file, myassigneeids.dta, contains assignee names and IDs based on this hand-check of u-turns with very similar names.