Talk:Global AbuseFilter

This page is for discussions related to the Global AbuseFilter page.   Please remember to: Sign posts using the four tildes (~~~~) Remain civil and polite during discussions. Place new text under old text (start a new post).   For older conversations you can see the archive index.

Archives Archive index

Special:AbuseFilter/72

Hi,

Could you please fix this filter so that it doesn't match "page" after any word ending with "y", e.g. "Larry Page"?

Thanks,

Orlodrim (talk) 20:24, 8 February 2019 (UTC)

@Orlodrim: Can you please reference who made the edit attempt. The filter isn't that naive, there are a number of components that would take place, so if we can at least mention the editor, then we can look at the issue. This is predominantly a filter that catches spambots, so it is very busy, and finding a false positive in among the stream of real p+ves can be a challenge.  — billinghurst sDrewth 01:45, 20 February 2019 (UTC)

Noting that I have made a tweak to the filter to look for a phrase termination, however, the above information would be good to allow for a re-test.  — billinghurst sDrewth 01:49, 20 February 2019 (UTC)

@Billinghurst:

Thanks for updating the filter. It works for the edit that was initially reported as a false positive, namely fr:Special:AbuseLog/2409751. However, due to normalization, "phrase termination" is basically "page termination", so it might prevent good detections too.

On the other hand, is the "y" really useful? I tried to understand why it was there, and I found more false positives: Special:AbuseLog/552157, Special:AbuseLog/552298.

Orlodrim (talk) 07:12, 20 February 2019 (UTC)

There was also a coding error in /72, which I have amended. In the second batch of errors, I will see what we can do about the first, the second doesn't concern me, that is just junk edits. Let us keep watching, it will never be error free.  — billinghurst sDrewth 11:13, 20 February 2019 (UTC)

Just to be clear, the main point of my previous message was that I suspect that your first fix could decrease the efficiency of the filter. For instance, Special:AbuseLog/551215 is no longer detected. The second half was a suggestion of another possible fix in case you decide to undo the initial one. Orlodrim (talk) 12:14, 20 February 2019 (UTC)

Undid the termination aspect. I am guessing that it was link to "y" as referring to "my", so we could be more specific in that regard.  — billinghurst sDrewth 21:29, 20 February 2019 (UTC)

ready steady girls stopped by spam filter

(cross-posted from https://www.wikidata.org/wiki/Wikidata:Project_chat)

I wonder if readysteadygirls.eu - a pretty good blog on European female 1960s pop singers imo - is stopped by the spam filter for actual abuse, or if it's caught by some "suspected porn" heuristic based on the domain name? Not a huge deal, but personally/anecdotally I've found this to be a well-kept and information dense website dedicated to female musicians. Moebeus (talk) 16:34, 19 February 2019 (UTC)

@Moebeus: I can try to look this up for you. Did this block an edit you tried to make? If so on which project, to which page, and when? — xaosflux ^Talk 18:06, 19 February 2019 (UTC)

@Xaosflux: I only noticed it when I tried to link an artist in Wikidata to an article about her on the readysteadygirls.eu site using the https://www.wikidata.org/wiki/Property:P973 ("described at URL"). I was not allowed to save with the UI displaying a red error box explaining that the domain was blocked by the spam filter. It happened yesterday. Thanks in advance! EDIT: The exact page I was trying to link to was http:// www.readysteadygirls.eu /rosalia/ 4544298600 but even the editor here won't let me save it ;) Moebeus (talk) 18:12, 19 February 2019 (UTC)

@Moebeus: ok I looked up your block, it is:

2019-02-18T18:18:26 Moebeus (talk | contribs) caused a spam blacklist hit on Rosalía (Q5394851) by attempting to add http://www.readysteadygirls.eu/rosalia/4544298600 http://www.readysteadygirls.eu/rosalia/4544298600.

This was not a Global Abusefilter hit, it is a global Spam blacklist hit. See the header at Talk:Spam blacklist for how to proceed. — xaosflux ^Talk 18:58, 19 February 2019 (UTC)

@Xaosflux: Aha, thank you for investigating! Moebeus (talk) 19:02, 19 February 2019 (UTC)