When using PATSTAT's table tls206_person, we notice that several character corruption can be found in columns person_name and person_address. This is also the case for other string cotaining tables such as  tls202_appln_title and tls203_appln_abstr. An analysis on person_address column corrupted records allowed us to distinguish a pattern on the corruption. And then to create a table to recover the (assumed) original characters.

•  Corrupted-original characters table (Excel file, 29Kb)
  (This table was built from table tls206_person column person_address records from April 2006 version, comments and corrections are very welcome)
   
• MySQL stored function correcting identified corrupted characters (text file, 4Kb)
   (To be used on MySQL)

Other programs:

• Transact SQL program correcting identified corrupted characters (text file, 5Kb)
   (To be used on MS SQL server)
• VBA program correcting identified corrupted characters (Basic file, 4Kb)
   (To be used on MS Access or Excel)