Token group types

clapf supports different token group types:

  • shared

All users share a common token set. This may be useful if you have plenty of users and they have similar email patterns or if you don't have enough disks to store many individual token sets.

If you want a shared database, then set group_type=0 in clapf.conf (this is the default).

  • merged

You may create a global token database with the initial training, then your users may customise it with their own messages. Clapf merges the two token set (global + user) at run time, and uses the merged set for determining probabilities.

It offers greater flexibility and may provide better accuracy, but requires more disk space to store the lots of tokens. If you add a new user later, the global database presents a default training set, and provides out of the box filtering.

If you want a merged database, then set group_type=1 in clapf.conf.