[Rspamd-Users] Could use some help with bayes classifier

Mathieu Marnat mathieu.marnat at sendinblue.com
Mon Jul 12 13:33:55 UTC 2021


First of all, thank you for providing this tool ! It is very thorough.
I would really like some help if anyone has time to check my case.

*Now, about my issue (config enclosed) :*
I am using Rspamd on a local VM.
I am training the bayesian filter using both spams and hams (around 400 of
each - and I do get Success : true while doing so). I use a separate
instance of Redis, where I can see that the "learned_ids" key is quite big
while "BAYES_SPAM_keys" and "BAYES_HAM_keys" have a serialized length of 4.
Are these keys supposed to get bigger ?

Also, I tried to use "learn_spam" for around 400 emails, and then scan one
of these emails (rspamc < email) which then gets me the following error in
the log : "bayes_classify: no tokens found in bayes database (20 total
tokens, 12 text tokens), ignore stats". Is it supposed to happen ?
When I try to analyse other emails not already learnt, I get the same error
message. Or sometimes not but I don't have any BAYES symbol showing up in
the result.

Am I missing something? Any help is appreciated!


-------------- next part --------------
A non-text attachment was scrubbed...
Name: rspamd_config
Type: application/octet-stream
Size: 110200 bytes
Desc: not available
URL: <https://lists.rspamd.com/pipermail/users/attachments/20210712/0d833563/attachment.obj>

More information about the Users mailing list