[Spambayes] Training Question - Status configuration
Tony Meyer
tameyer at ihug.co.nz
Tue May 9 04:42:12 CEST 2006
>> We recommend not training on all messages, but rather using a
>> mistake-
>> based training approach. Train any messages that are incorrectly
>> classified and any messages that are unsure.
>
> When I double click on the tray icon I see what's been
> processed since the last time I used the train button.
> I don't see a way to select only a certain number or
> variety of processed emails.
You have four options for each message, defer (i.e. do nothing with
it), ham (train it as good mail), spam (train it as bad mail), and
discard (do not train it). This allows you to do any training regime
that you like.
> I can click the defer options on the good emails and
> be sure the SPAM was listed as SPAM before clicking
> on the training button but the next time I check to
> see what has been processed, all the good emails
> previously set as defer return listed as HAM & I have
> to go through all of those which is a lot of redundency
> on my part.
>
> So I don't see a way of bypassing the emails properly
> classified as HAM, they still get "trained".
Why not use discard?
>>> there's no global option to list everything
>>> as defer & then select the Spam & equal #'s of HAM
>>> for training purposes.
>>
>> Yes, there is. These are the "Default training for ham", "Default
>> training for spam", and "Default training for unsure" options. They
>> are on the advanced configuration page.
>
> Those options appear to be set by default, I don't
> think I changed them.
Yes, every option that SpamBayes has has a default value.
[...]
> Perhaps you're suggesting I list everything as unsure
> and then correct that designation after the email has
> been processed?
My suggestion is that you only train on mistakes and unsures. So, by
default, it would make most sense to discard ham and spam, and defer
unsures. Then you need to correct the value for all the unsures and
any mistakes, and that's it.
=Tony.Meyer
--
Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes.
http://www.massey.ac.nz/~tameyer/writing/reply_all.html explains this.
More information about the SpamBayes
mailing list