Statistical Spam Filter for Comments

What is Statistical Spam Filter for Comments?

Statistical Spam Filter for Comments is a submodule of the Statistical Spam Filter (SSF) module that is part of the same distribution. It makes use of the functionality of the SSF module to classify incoming comments as spam or ham (non-spam). Ham is delivered to the Published comments or to the Unapproved comments (depending on permissions).

Spam is delivered to the Spam comments box, while messages, for which the SSF cannot classify the comments as ham or spam, are delivered to the Unapproved comments where a human operator can assign these comments to the Published comments or the Spam comments.

How the filtering works

Before the SSF module can reliably classify comments as spam or ham, it needs to learn for a couple of comments which messages are spam or ham. This learning takes place when a human operator decides to assign Unapproved comments to the Published comments or to the Spam comments.

Although SSF learns the difference already within a couple of comments of both types it is advisable to not allow the publishing of comments without human approval during the learning phase, even though you can always re-assign comments from published to spam or vice versa by hand (SSF learns from those actions too).

When the SSF module has learned to tell the difference between ham and spam you can switch off the approval of comments by a human operator in the permissions (Skip comment approval).

Spam filtering and blocking works right out of the box after installing both the Statistical Spam Filter module and Statistical Spam Filter for Comments module and enabling both modules with the default settings.

Settings

The settings are two parameters that determine the gap between the classification of ham and spam messages. The gap may seem large, but clear spam and ham messages tend to be classified to the extremes of the classification range.

However, when too many comments wind up in the Unapproved comments it is worthwhile to decrease the gap somewhat.

As extra functionality the module can send e-mail notifications when unapproved comment are posted that are waiting for approval. However, e-mails are sent in the HTML format. You will need a module such as Swiftmailer to send HTML e-mails.