use dewiki_p;

  #SELECT COUNT(DISTINCT page_title, cl_to, cl_sortkey_prefix)

  SELECT REPLACE(page_title, '_', ' '), cl_sortkey_prefix

   FROM page

   join categorylinks on page_id=cl_from

   WHERE cl_sortkey_prefix != "" and

   page_namespace = 0 and

   cl_to not like "Wikipedia:%" and

   LOWER(cl_sortkey_prefix) not like LOWER(REPLACE(page_title, '_', ' ')) and 

   cl_sortkey_prefix not regexp '^([^[:print:]]|[[:alnum:]]|[./ \--–()]|(#|[0-9]):+[0-9])+$'

#limit 300

;