How to get Familiennamen from a German corpus with 90% Precision
and unknown Recall.

1. Get all bigrams with DDC for the query:
   "=firstname.upr :UPPER\"
where  firstname.upr is a file with first names, without a special list 
"bad_ger_names.txt", which contains very  ambigious in German  words
like "Nur", "Mark", "Tod" and so on.
The second column of this file contains possible surnames.

2. Convert the list to lower case;

3. delete all lines, where the surname consists only of vowels or consonants.

4. sort and delete duplicates of lines. Now we have a list of pairs like "Name Surname",
where each pair is unique.

5. Delete the first column in the table, sort, and delete all surnames
which occur only once. Now we have a list  of surnames which occur 
at least two times with different(1) first names.

6. Delete all surnames, which can be recognized by English or German morphologies.

7. Delete all surnames which terminate with "str" and "ung".

 