KES is the Kbin Enhancement Suite, a userscript/extension for (k/m)bin that provides a variety of customizable tools for users.
The last minor version included an initial attempt at a spam filter. This was chiefly done to address the low-hanging fruit of spam and scrub the most persistent pharma ads, etc. The approach was similar to that used by Ublock Origin or Steven Black's hosts file, in that it was a monolithic list of filter rules.
This was alright as a stopgap measure, but to the surprise of no one, the types of spam that continued to appear were innumerable in variety.
In an effort to have some filtering rather than none, the feature also went against the KES dictum about giving users choice to tweak their own settings: it was an all-or-nothing filter.
This update introduces v2 of spam filtering. The old logic has been retired, but may make a comeback at some point. A new sidebar page titled "Spam" has been added, and this will be a central place for anti-spam features.
The first of these efforts is a new filter that exposes the following options:
- Hide posts from very new accounts
Users commented that there is no "gating" mechanism with registration, causing bogus accounts to be constantly recreated as a form of ban evasion. If you find that very new accounts have a high tendency to be spam, you can tick this setting to remove posts from very new users from the feed.
This does not block the user outright, merely hiding the post. This effectively gives some "break-in" time for the user to prove themselves. After a certain threshold, the user will fall outside of the "new account" window; if they continue to engage in spam-like behavior at that time, they should be caught by the other ban filters below.
- Hide posts with abnormally low relative rank
Users with aggregate posting activity that does not necessarily resemble spam at first glance, but which periodically post spam/off-topic/controversial threads in the wrong communities, have a high likelihood of being astroturfers. This option hides such posts from the feed if the relative vote weight is egregiously outside the norm. Like the above, it does not block them outright.
- Block users with spam-like activity
This feature blocks users outright if they engage in spam, mass posting, and other robotic behavior. It is less permissive than v1 of the spam filters, which failed to catch a lot of spam that was not explicitly blacklisted in the filter list. This new approach is more generic and does not need constant updating of the spam filters.
By default, all three options are enabled. I find this combination gives the best results so far. Note that infinite scrolling should be enabled for best results, and it may take a little while when navigating to a magazine for the results to be filtered.
Give it a try and see if this negates the spam issue or improves your experience. What other filter options would you like to see added?
The remainder of the changelog follows:
Add-ons
Filter advertisements (@shazbot)
Location: Spam > Filter advertisements
Updated filtering logic to v2, as described in the preamble of this post.
Collapse pinned posts (@shazbot)
Location: Threads > Collapse pinned posts
This simple feature groups together pinned posts on the magazine index and collapses them by default. Click the toggle area at the top of the magazine index to expand pinned posts. This is especially useful on magazines with many pinned posts that are not regularly archived, such as /m/kbinmeta.
API
Added the helper function isThread() to the API. This returns true if the current window location is a thread inside of a magazine. This can be used to abort if the feature is intended to only apply on the thread index. In the future, a more expanded function will be provided that returns the type of page from a list of enums.
If your main goal is exporting your magazine subscriptions between accounts across instances, may I suggest trying EXIT tool. If you are looking for more complex export settings (friends/favorites?), unfortunately, only subscriptions are supported at this time.