[quote user="amoradell"]
But I have to download mails from my pop servers.
[/quote]
Spamhalter is a text classifier and needs the whole message to decide if it's spam or ham.
[quote user="amoradell"]
[...] often, email address and subject is really sufficient...
[/quote]
The email adress contains a lot of random stuff. Only parts of it can be helpful.
A decision based on such a small text corpus (subject line and email adress) means a
higher probability of misclassifications and is a little bit shaky.
[quote user="amoradell"]<p>But I have to download mails from my pop servers.</p>
<p>[/quote]</p><p>Spamhalter is a text classifier and needs the whole message to decide if it's spam or ham.</p><p>[quote user="amoradell"]</p>
<p>[...] often, email address and subject is really sufficient...</p><p>[/quote]</p><p>The email adress contains a lot of random stuff. Only parts of it can be helpful.</p><p>A decision based on such a small text corpus (subject line and email adress) means a</p><p>higher probability of misclassifications and is a little bit shaky.
</p><p>&nbsp;</p>