You can use sa-learn tool but it's not recommended because it requires advanced knowledge for this tool and wrong utilizing can cause serious consequences for spamassassin.
Here is one tip for your case:
You can use provided solution for manual utilizing of sa-learn
> > sa-learn --spam c:\program files\swsoft\plesk\mail servers\mail enable\domains\domain.com\mailroot\mailbox\inbox
This command puts learnt bayes files to
C:\Documents and Settings\Administrator\.spamassassin.
But personal mail box SA configuration files a stored at
C:\Program Files\SWsoft\Plesk\Additional\SpamAssassin\SpamFilterUserConfigsPath\
[email protected]\.spamassassin
so you should copy them manually.
Commonly learn algorithm can be described as the following:
1. Copy from C:\Program Files\SWsoft\Plesk\Additional\SpamAssassin\SpamFilterUserConfigsPath\
[email protected]\.spamassassin to C:\Documents and Settings\Administrator\.spamassassin.
2. sa-learn --spam c:\program files\swsoft\plesk\mail servers\mail enable\domains\domain.com\mailroot\mailbox\inbox
3. Copy from C:\Documents and Settings\Administrator\.spamassassin back to C:\Program Files\SWsoft\Plesk\Additional\SpamAssassin\SpamFilterUserConfigsPath\
[email protected]\.spamassassin
Note: This feature should be performed only under human monitoring and control.
To browse generated bayes database please use:
C:\>sa-learn --dump --dbpath "C:/Program Files/SWsoft/Plesk/Additional/SpamAssassin/SpamFilterUserConfigsPath/
[email protected]/.spamassassin"
Spamassassin under Plesk for Windows has more poor management interface than Plesk for Linux because Plesk for Windows supports several mail servers and every of them have quite different mail message structure and full mechanism is still not completed to the end.