You can already exclude pages containing a string, very easily. Are you saying that this kind of regular expressions you're looking for don't work with AbuseFilter? (I don't remember what subset of regex we use exactly.)

Anyway, your approach seems wrong: you can't determine context reliably with regex, you should tag such edits adding words you don't like and then check them manually.

(In reply to comment #1)

You can already exclude pages containing a string, very easily. Are you
saying
that this kind of regular expressions you're looking for don't work with
AbuseFilter? (I don't remember what subset of regex we use exactly.)

Anyway, your approach seems wrong: you can't determine context reliably with
regex, you should tag such edits adding words you don't like and then check
them manually.

:The logic is exemption is needed only in a particular line in particular page.Use anywhere else is abuse and can be disallowed strait away.

:This will increase work for abuse filter only initially.Once the people know they can not cheat the system easily they will automatically desist.Once people desist lesser work for filters,filter managers and patrollers.

:Tagging only continues patrollers work.Wiki's like en wiki has large manpower available to do that.Most of the wikis in wikimedia umbrella are smaller wikis and we have less man power .If we save our time on patrolling our editors will have more time at disposal for content contribution

Mahitgar, you can achieve that easily by changing the condition to

( ( YOURCONDITION ) & article_text != '13 June' )

(In reply to comment #3)

Mahitgar, you can achieve that easily by changing the condition to

( ( YOURCONDITION ) & article_text != '13 June' )

Yes I know, this exempts the complete page.

Thanks

One can already write a filter that checks for something like

("word" in added_lines) & ! ("<!-- hidden tag -->word" in added_lines)

and similar constructions, where editors agree in advance to use a particular marker to indicate that a word or phrase is okay. That seems like a problem for filter writers / article editors to decide and implement. Beyond that, I'm not sure what the actionable request is here.