Wikipedia:Bots/Requests for approval/ClueBot NG/Old False Positives

Old False Positives

Diff	Comment	Reason/explanation/fix/discussion etc.
	The bot undid my interwiki link for the Template page. 78.72.250.55 (talk) 13:41, 18 November 2010 (UTC)[reply]	A few of the other false positives above have also been related to interwiki links - it's a dataset issue. As the dataset grows, this should be fixed. Crispy1989 (talk) 16:08, 18 November 2010 (UTC)[reply]
Edit	The bot undid an IP user's good faith edit, [1], a perfectly fine insertion of a word for clarification. I have reverted back and intend to post on the IP talk page that it was a mistake. I realise it might be harder to distinguish between such edits and some forms of vandalism, but I imagine minor edits of this type are common and we don't want to deter casual editors. ChiZeroOne (talk) 16:48, 18 November 2010 (UTC)[reply]	This is one of those weird ANN things that will definitely be fixed by a larger dataset. The word "accidental" does not appear in the Bayesian database at all, so the statistical properties of the message must just be falling into a strange gap in the training set. Crispy1989 (talk) 16:50, 18 November 2010 (UTC)[reply]
[2]	Bot undid an edit on [3] from SML/Texting to SMS/Texting. SML documentation provides no mention of IDC, and the person who added SML/Texting was someone who has previously vandalized. As SMS is the only topic relating to the definition that follows the SML/Texting entry ("Idc" or "I.D.C." is an acronym for "I don't care.") I made the change to SMS/Texting. Phoenixkin (talk) 18:27, 18 November 2010 (UTC)[reply]	The dataset had not previously seen "SMS" or "SML" in all caps, and as such, treated it as shouting. Crispy1989 (talk) 01:18, 19 November 2010 (UTC)[reply]
Edit	Bot reverted edit that wasn't vandalism, but fancruft. [4] I reverted the bot edit to let the editors decide. --Confession0791 talk 21:28, 18 November 2010 (UTC)[reply]
	I just added a note that QuickTime is required to hear the Start-up and Chimes Of Death in MacTracker. --216.96.2.184 (talk) 22:53, 18 November 2010 (UTC)[reply]
	Bot reverted an edit where I added the relevant quote to a reference. Quote was buried in lengthy source text therefore quote in reference was necessary & useful
	Added a new link to the recently-released English version of the product website, but was reverted.
Edit	I don't believe that this should have been reverted. The IP was simply trying to say that the actress is now in another soap and it got reverted as possible vandalism. --5 albert square (talk) 09:38, 19 November 2010 (UTC)[reply]
	I seem to have gotten credit for several vandal edits from this IP address on subjects which I have never written about-footballers and the Fritzl family and so on. The edit on "The Sparrow" was legit.--83.70.226.255 (talk) 16:09, 19 November 2010 (UTC) Also will someone please tell Jimmy wales the multimillionaire megalomaniac philanthropist to take his smirkung begging face off every page ![reply]
	The edit by anonymous user 24.215.26.57 was perfectly legitimate. It appears that the bot was triggered by the facts that the edit was made by an unregistered user, combined with the added material being in full caps (a stylistic gaffe, to be sure, but one that can easily be corrected).—Jerome Kohl (talk) 16:37, 19 November 2010 (UTC)[reply]
	I was pointing out that the legend of the swan song was a misconception, as mute swans do not "sing right before their deaths". The bot reverted the one-word edit. 192.12.88.68 (talk) 03:01, 20 November 2010 (UTC)[reply]
Edit	I wonder why this one triggered the bot. It was a pretty clear similar addition to what was already on the page. ? Melodia Chaconne ? (talk) 23:07, 15 November 2010 (UTC)[reply]
Edit	Or this one, which seems like a legitimate addition to the Languages list for the article. (I don't read this language, but I was able to locate the article.) Cynwolfe (talk) 00:16, 16 November 2010 (UTC)[reply]
List of Representatives from North Carolina	My edit was misconstrued as nonconstructive. I was trying to put up a cleanup tag on the page List of Representatives from North Carolina, and not having enought experience doing that, I did it wrong. 71.91.99.47 (talk) 01:52, 16 November 2010 (UTC)[reply]	This is mostly an issue with dataset completeness (not many/any instances of incorrectly added templates), but it probably would not have been classified as vandalism if that IP had not vandalized in the past. Crispy1989 (talk) 02:08, 16 November 2010 (UTC)[reply]
Edit	This [5] reads like a content dispute where the bot should not be taking sides. Philip Trueman (talk) 08:32, 16 November 2010 (UTC)[reply]	I don't see any reason this edit would be classified as vandalism. It must be a dataset completeness issue. Crispy1989 (talk) 12:46, 16 November 2010 (UTC)[reply] There's a change of "facebook" to "myspace", and that might have been a contributing factor, since I presume vandals sometimes spam social site links. —  HELLKNOWZ  ?TALK 12:50, 16 November 2010 (UTC)[reply]
[6]	[7]. Trying to improve the page. --130.233.79.47 (talk) 08:34, 16 November 2010 (UTC)[reply]	Just asking, but has the problem of "i.e." being treated as ultra-short sentences been fixed? Philip Trueman (talk) 10:50, 16 November 2010 (UTC)[reply] I thought I fixed it, but seemingly not. There's no other reason this edit should have been reverted. To fix it before, I just removed all instances of "i.e." before processing. I just changed it to instead replace it with "Ie". Crispy1989 (talk) 12:37, 16 November 2010 (UTC)[reply] I checked for that. Actually it was myspace replaced with facebook, so the bayesian score would be influenced by the addition of "facebook", but "facebook" isn't even in the bayesian database (close to 50/50 vandalism/constructive occurrences). There are a few words that are in the database, like "rumors", "very", and "happy", but none of these have particularly high scores - definitely shouldn't be high enough to cause a false positive. There must just be some statistical property of the edit that fell into a gap in the ANN's training set. Crispy1989 (talk) 12:54, 16 November 2010 (UTC)[reply]
Edit	Nothing to do with me really but this edit doesn't look like vandalism to me. Rambo's Revenge (talk) 20:04, 16 November 2010 (UTC)[reply]	The bot was treating what was inside there as a quote, and text inside quotes isn't processed (other than counted). Combined with the multiple other recent warnings, it was enough to trigger it. Crispy1989 (talk) 21:30, 16 November 2010 (UTC)[reply]
	That what they do is vandalism. The Vlachs are not Romanians, they are ethnic group recognized by the Constitution! In census their language is recognized as Vlach! So, we must to respect that. Thanks!	The edit you're referring to looks like vandalism to me. It's not obvious vandalism, but removing all mention of "Romania" from an article where it is pertinent usually constitutes vandalism. The bot is capable of catching non-obvious vandalism such as this. Crispy1989 (talk) 16:06, 18 November 2010 (UTC)[reply]
Shawn Johnson	There was no vandalism. This is an automated process, so not sure why my edit was tagged and reverted. I wanted to keep working on the article!
	Re Tropaeum Traiani cluebot NG revert 10.18am 20 November 2010 I simply added the Latin text of the memorial underneath the section heading "Legionares Memorial" The text is from note2 source, this is not "vandalism".
Edit	I was adding a source that is already in the article! Followorders (talk) 21:43, 20 November 2010 (UTC)[reply]	False positive due to Bayesian keyword. Will be mitigated by complete parser and fixed by improved dataset.
	The stupid bot deleted my entry! MAKE IT STOP, MOM!!! 173.59.219.49 (talk) 22:31, 20 November 2010 (UTC)[reply]	Not a false positive. The edit is vandalism.
Edit	Correcting name of a band from "Wakey Wakey" to "Wakey!Wakey!" but the system thinks it knows best...	Sole addition of two (non-consecutive) exclamation points. Might be fixed by improved dataset, but very rare edge case anyway.
Edit	Looks like the user was adding the "Smoky Bacon" flavor of Pringles chips and was nailed for it	Since the user did not add a space or comma, it was treated as a very long nonsense word. A full parser may help this, as could a very extensive dataset including examples of this kind of well-intentioned error.
http://en.wikipedia.org/w/index.php?title=Harry_Potter_and_the_Deathly_Hallows_(film)&diff=prev&oldid=398006837	Dude, WTF? You restored an every-word-has-a-link vandalism.	Definitely fixed by larger dataset. With the current dataset, all mass removals of links have been vandalism. Clearly this is not always the case.
Edit	Previous version by 70.26.181.136 had corrected errors and improved the section with valid text. Your edit removed the addition and restored the typo corrected by the editor.