User talk:Emijrp/Good and evil list.css

Suggestions[edit]

Suggested Bad Words[edit]

Words or phrases that are associated with vandalism.

Suggested Test-edit Words[edit]

Words or phrases that are associated with test-edits.

Suggested good Words[edit]

Words or phrases that are indicators of a good edit.

I said before that quotation marks are generally good; at least 2 false positive edits contained them [1] [2]. I'd add that brackets are also generally good too. Sole Soul (talk) 06:02, 4 March 2010 (UTC)[reply]

Done Thanks! I added double-quotes to the good list, along with bracketed http links. Also, tranclusions and wiki-links are already given good points. Tim1357 (talk) 02:40, 5 March 2010 (UTC)[reply]

Capitalized first letter word in the middle of a sentence. Reducing the negative score for "Love" and "Dude" is missing the point of why they are false positives. They are false positives because they are proper nouns. This will likely happen again with words like "Fuck" as a proper noun. Other examples of FPs [3] [4] [5] [6] [7]. Sole Soul (talk) 22:09, 21 March 2010 (UTC)[reply]
The word "plot". Sole Soul (talk) 18:43, 5 April 2010 (UTC)[reply]
The word "character". Sole Soul (talk) 18:51, 5 April 2010 (UTC)[reply]
I suggest the following regex to detect proper nouns (which are frequently appear as FPs) in the middle of sentences (with the help of a regex cheat sheet):

\s[a-z]{2,},?\s[A-Z][a-z]+

It matches the following FPs edits: [8], [9], [10], [11], [12]. It matches other FPs indirectly [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23]. Sole Soul (talk) 06:42, 26 April 2010 (UTC)[reply]

Useful pages[edit]

Useful pages for new regexps:

emijrp (talk) 16:18, 28 February 2010 (UTC)[reply]

User:Lupin/badwords Doing... Tim1357 (talk) 21:07, 28 February 2010 (UTC)[reply]
User:ClueBot/Source#Score_list Done Tim1357 (talk) 21:07, 28 February 2010 (UTC)[reply]

Sole Soul (talk) 20:50, 28 February 2010 (UTC)[reply]

Info about regular expressions policy[edit]

Hi. Thanks for working on this list. When a pattern is a nice (for example: [[a link]]), points must be positive (+1, +2, +3...). When it is bad (for example: fuck), negative (-1, -2, -3). For reverting always (for example: a vandalism campaign attack) must be -9999. I'm going to add this info to .css page. Regards. emijrp (talk) 19:47, 22 February 2010 (UTC)[reply]

Cool, thanks for the update. What is the threshold score for when the bot would revert an edit. I say maybe 5. What do you think? Tim1357 (talk) 21:48, 22 February 2010 (UTC)[reply]

If score<-4, then bot reverts. If score<0 and score>=-4, it depends on the amount of text inserted by the user, because, a vandalism density is calculated. For example, an anonymous user can insert the word "shit" with 1 KB of text, because of "shit" can be a cite, so bot doesn't revert. But if the anonymous user inserts 15 bytes and the word "shit", it is a blatant vandalism. You can see it at line 90. Of course, these parameters can be modified, but, my experience in Spanish Wikipedia says that it works fine. emijrp (talk) 21:58, 22 February 2010 (UTC)[reply]

That sounds about right. I updated the sandbox, using your changes. You changed '.' to '[a-z]'. The intention of that regexp was to catch repeating strings, such as 'PENIS PENIS PENIS'. The regex dosent work if it is not a dot. Tim1357 (talk) 22:16, 22 February 2010 (UTC)[reply]

P.S. Would you mind copying and pasting the sandbox list to the main page for me? Tim1357 (talk) 22:20, 22 February 2010 (UTC)[reply]