« first day (1675 days earlier)      last day (2144 days later) » 

9:00 AM
oh, I'm looking at the code, you use the levenstein distance
 
just in one or two of the rules, the main bulk is just regex
 
do you have a dataset with the positives and negatives?
cause we could implement a NLP AI
 
the link to Metasmoke up in the right-hand room description
 
what's the accuracy of the smoke detector? (sorry if I'm doing too much questions xD)
 
9:01 AM
@Neoares yeah, we had a machine learning hotshot turn up here around a year ago but he disappeared after delivering a deep learning solution which didn't do anything
 
I think you can have either a good accuracy or a good recall, but not a good F1-score
 
that is not to say it could never work, if you have those skillz we'd be very happy to try to set up a few experiments
 
[ SmokeDetector | MS ] Bad pattern in URL answer: Are Adobe Typekit fonts "free" to use? by alx1xla on graphicdesign.SE
 
A J
@Glorfindel \o/
 
9:02 AM
fp- by tripleee
 
well I'm not a NLP expert... I work with images as an input for my NNs
 
@Neoares the amount of completely predictable spam on the sites we monitor skews things substantially, we could get very close to 100% on those with no false positives but we try to cast a wider net
recently some trolls have turned up and chasing those has been a big ole game of whack-a-mole which detracts from the work on the spam filtering and brings a fair deal of FPs when we try to predict what they will come up with
 
I dunno, but if you do an AI, you should use the feedback provided by the human as a feedback for your net
I don't know if that was being done in your NLP experiments
 
the Metasmoke database already contains feedback for 100k+ posts
 
oh, nice
 
9:06 AM
but it's heavily skewed towards spam, so for training you should probably get samples of non-flagged posts from a separate source
or just run your own instance for a day or two and collect posts into a local repository
 
[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, bad pattern in URL body, blacklisted website in body, link at end of body, +2 more: Phen375 - Quick & Natural Way To Lose Extra Fat! by Phen375 on meta.SE
 
page not found
 
I have I think on the order of 10k of posts which were not flagged which I collected for something like that over around a year I think but it's been a while since I last did a run
 
tpu- by tripleee
 
@Neoares it was deleted because of autoflags, you were too slow
metasmoke.erwaysoftware.com/post/123944 shows the Metasmoke archive of this post and tells you it was deleted within 20 seconds
 
9:09 AM
tp by PeterJ
 
oh god... this metasmoke is amazing
all the thing in general
 
[ SmokeDetector | MS ] Username similar to website in answer: Emulation on a Linux PC by iotify on raspberrypi.SE
fp- by tripleee
 
good one
false positive?
 
Yep, that's what the fp means.
 
there is a website called iotify...
it wasn't a false positive
 
9:12 AM
the comments below the post on the site from a site moderator seem to confirm my conclusion but one of the things we have this room for is debating these corner cases and trying to find support from site policies for our decisions
 
It specifies that they're affiliated with the site and it actually answers the question, so it's acceptable.
 
maybe it should be naa though, it doesn't really answer the question
("not an answer")
 
IusetifywithsomebuzzwordtoattractvcfundingbecauseIamatalentlesshackthathasnoothe‌​rcareerprospectstify
 
9:16 AM
oh god... who wrote the smoke detector?
 
those regex are big
 
@Neoares Many many many people did
 
I hope so
 
@Neoares It's a collaborative process over the past few years.
 
9:17 AM
particularly people listed here: charcoal-se.org/people
 
@Mithrandir lol
 
@Mithrandir all the blacklist and watchlist commands generate github comitts by smokey itself
 
9:18 AM
[ SmokeDetector | MS ] Username similar to website in answer: Sharing Bitmap via Android Intent by kjs566 on stackoverflow.com
 
@Magisch I know, it's just amusing
 
haha
 
[ SmokeDetector | MS ] Link at end of answer, pattern-matching website in answer, potentially bad keyword in answer: How to copy a local Git branch to a remote repo by Priya Priya on stackoverflow.com
 
smokey grew with the ambitions of the people running it
 
9:19 AM
tpu- by Zoe
 
these days we're automatically flagging posts and creating records and even tagging domains
 
what's the U in tpu
 
blacklist user
 
Zoe
@Neoares blacklist user
 
[ SmokeDetector | MS ] Username similar to website in answer: Get JavaFX View as Object? #Update2 by GJCode on stackoverflow.com
Recovered from websocket._exceptions.WebSocketConnectionClosedException: socket is already closed.
fp- by tripleee
 
9:22 AM
there's a lot of different and funny questions in here
 
most of all this is an archive of filth
you will find no more condensed pile of literary feces on all of SE
 
that explains the smell when I joined the chat
 
9:23 AM
!!/allnotifications 11540
 
You will get notified for these sites:
ai.stackexchange.com, conlang.stackexchange.com, literature.stackexchange.com, puzzling.stackexchange.com, scifi.stackexchange.com
 
it gets compensated with beauty and health spam though
 
and success, let's not forget the success
 
tp by Mithrandir on SEO text in hidden element [MS]
 
9:24 AM
what came of Undo's attempt to sign up with the Illuminati?
 
unfortunately the spam here is much less appetizing then the regular canned luncheon shoulder meatloafs
 
!!/test 1bandsaw.com
 
> Blacklisted website in body, blacklisted website in title
----------
Title - Position 1-13: 1bandsaw.com
Body - Position 1-13: 1bandsaw.com
 
9:28 AM
!!/test-a telorevive
 
> Would not be caught as an answer.
 
!!/watch telo\W?revive
 
@tripleee Added telo\W?revive to watchlist
 
!!/test Navari Testo
 
> Would not be caught as a post, title or username.
[ SmokeDetector ] SmokeDetector started at rev bcd0fd1 (SmokeDetector: Auto watch of telo\W?revive by tripleee --autopull) (running on teward/Solar Flare)
Restart: API quota is 16340.
 
9:29 AM
!!/blacklist-keyword phen375
 
@tripleee Blacklisted phen375
 
!!/blacklist-keyword Navari\W?Testo
2/2 in 8 days; may be part of an upcoming campaign (see here)
 
@Mithrandir Blacklisted Navari\W?Testo
 
looks like a typo of Narvi though
or maybe the other way around
 
Restart: API quota is 16327.
[ SmokeDetector | MS ] Username similar to website in answer: Sharing Bitmap via Android Intent by kjs566 on stackoverflow.com
[ SmokeDetector | MS ] Username similar to website in answer: how to save and share bitmap in android by kjs566 on stackoverflow.com
Restart: API quota is 16283.
[ SmokeDetector | MS ] Username similar to website in answer: How to attach a Bitmap when launching ACTION_SEND intent by kjs566 on stackoverflow.com
fp- by ByteCommander
 
Zoe
9:37 AM
@SmokeDetector fp- Duplicated answer, flagged
 
Recovered from ssl.SSLError: [SSL: BAD_LENGTH] bad length (_ssl.c:2130)
[ SmokeDetector | MS ] Blacklisted website in body, body starts with title and ends in URL, link at end of body: Spectrum CBD case of preventing ageing by veriw on askubuntu.com
tpu- by PeterJ
 
@tripleee Undo's Illuminati escapade: here, here, and here.
 
Restart: API quota is 16165.
 
so he chickened out in spite of their very credible response?
!!/blacklist-keyword spectrum\W?cbd
 
@tripleee Blacklisted spectrum\W?cbd
 
9:47 AM
2/2 bending the rules
 
CI on 2510c71 succeeded. Message contains 'autopull', pulling...
Restart: API quota is 16139.
Restart: API quota is 16130.
[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +2 more: freemp3download.club/blicky-da-blicky-22gz-mp3-song-download/ by jamesldsan on puzzling.SE (@Mithrandir)
[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, bad keyword in title, body starts with title and ends in URL, link at end of body: Vital X9 Male Enhancement by williahols4 on drupal.SE
[ SmokeDetector | MS ] Few unique characters in body: NOT WORKINGNOT WORKING by Helhior on stackoverflow.com
tpu- by Mithrandir on Vital X9 Male Enhancement [MS]
tp- by ByteCommander
 
!!/blacklist-website-force mumyfood\.com
 
@tripleee That pattern looks like it's already caught by Bad ns for domain in answer and Bad ns for domain in body; append -force if you really want to do that.
@tripleee Blacklisted mumyfood\.com
Restart: API quota is 16069.
 
in Charcoal Test, 1 min ago, by SmokeDetector
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, bad pattern in URL body, blacklisted website in body, blacklisted website in title, +4 more: rapidtone-reviews>>>> order4healthsupplement.com/rapidtone-reviews/ by Sultensem on superuser.com
this wasn't reported here, was it?
 
tpu- by PeterJ
 
9:57 AM
doesn't look like it, no
 
it's been a month since the latest troll post with that pattern, should we remove it?
 
[ SmokeDetector | MS ] Blacklisted website in body, pattern-matching website in body: Why not to snap up rs 3 gold with up to $10 cash coupon? by ESO2017123 on askubuntu.com
tpu- by grgarside
 
@tripleee +1
 
[ SmokeDetector | MS ] Luncheon meat detected: merge csv files python by Olivera Tancheva on stackoverflow.com
fp- by paper1111
 
10:05 AM
!!/unwatch ^umar$
 
@tripleee Removed ^umar$ from watchlist
[ SmokeDetector ] SmokeDetector started at rev f87e6c3 (SmokeDetector: Auto unwatch of ^umar$ by tripleee --autopull) (running on teward/Solar Flare)
Restart: API quota is 15987.
[ SmokeDetector | MS ] Bad NS for domain in answer, potentially bad keyword in answer: How does Xbox One gamesharing work when more than two accounts are involved? by rkjakhardesigns on gaming.SE
[ SmokeDetector | MS ] Luncheon meat detected: how get source code apk on linux by magic man on stackoverflow.com
[ SmokeDetector | MS ] Username similar to website in answer: What exactly is an AWS "Edge Location"? by AWS Clouds on stackoverflow.com
 
Zoe
@SmokeDetector disclosed
 
naa- by Zoe
@iBug Your flag conditions were manually disabled by @System: Total accuracy must be >= 99.75%.
 
@SmokeDetector of course a Kali Linux user... :facepalm:
 
J F
10:24 AM
@Magisch * spreadable canned luncheon shoulder meatloaves
 
naa- by tripleee
 
J F
\o/ Marshal on Graphic Design
 
fp by tripleee
 
probably fine actually, several of the other answers are quite similar
 
Restart: API quota is 15700.
 
10:42 AM
@JF congrats!
 
O.o where'd our tags go
@SmokeDetector This can't keep happening @iBug - you're going to need to find something that's consistently over the accuracy threshold.
 
2 hours ago, by tripleee
room topic changed to Charcoal HQ: Where diamonds are made, smoke is detected, and we break things by developing on production. For more information, see https://charcoal-se.org/about and of course https://metasmoke.erwaysoftware.com (no tags)
 
:( I like our tags
room topic changed to Charcoal HQ: Where diamonds are made, smoke is detected, and we break things by developing on production. 98.5k true positives and counting. For more information, see charcoal-se.org/about. Handy links: metasmoke.erwaysoftware.com, github.com/Charcoal-SE. [dev-on-prod] [panic-driven-development] [plastic-knives] [programmatic-crowbars]
 
A J
@ArtOfCode you mean "liked"?
 
no I mean like :P
 
10:46 AM
[ SmokeDetector | MS ] Bad keyword in body, blacklisted website in body, link at end of body, link following arrow in body, pattern-matching product name in body, +2 more: How can Naturnica function? by Naturnica on superuser.com
tpu- by PeterJ
 
A J
@ArtOfCode what do you mean like? or what do you like mean? :P
 
Could we shorten metasmoke.erwaysoftware.com, github.com/Charcoal-SE. to metasmoke and GitHub?
Waffles
 
No, links don't work like that in room descriptions.
And you need the https
 
Oh ok - shame
Would have saved some space
 
[ SmokeDetector | MS ] Potentially bad keyword in answer: How to restore corrupted InnoDB tables? by Alfie Jin on dba.SE
[ SmokeDetector | MS ] URL in title, bad NS for domain in body, bad NS for domain in title, bad keyword in body, bad keyword in title, +4 more: www.nutritionsofhealth.com/myvigra/ by user122750 on graphicdesign.SE
 
10:54 AM
@ArtOfCode FR to MS: Show an overall accuracy on all flag conditions page.
 
@iBug [status-declined], use the SQL Explorer query. Too heavy to do on every page load.
 
@SmokeDetector f
@SmokeDetector k
 
Restart: API quota is 15494.
 
!!/watch megainfomatix\.com
 
@Glorfindel Added megainfomatix\.com to watchlist
Restart: API quota is 15481.
 
11:00 AM
@ArtOfCode Well, good to have that.
 
that's why it exists, so that we don't have to do it in MS
 
@SmokeDetector I think this was spam: see the prior history of "wetechtuts" and the usernames involved: metasmoke.erwaysoftware.com/…
@ByteCommander ^
(I custom mod-flagged that post)
 
@ArtOfCode Or cache the routinely calculated result and display it?
 
@iBug then you don't get the number updated when the condition is updated and people get confused
one of those things that's almost certainly possible, but not worth doing when we have an acceptable alternative
 
fine.
I think there's huge room for improvement by pre-indexing the DB.
 
11:09 AM
@user6655984 it also doesn't answer the question; the OP wants to know how to do this with more than one friend
 
So in addition to using existing posts, we have an index like "Site Xxx, at weight xxx, there are x TPs and x FPs"
 
Restart: API quota is 15376.
 
Then we only need a cron job to update that index regularly.
 
@iBug PRs welcome :)
 
11:15 AM
@ArtOfCode If only I have the ability :(
Anyway, it's a good idea, isn't it)
 
you want to recalculate a quotient on a regular schedule?
 
Isn't it what we're already doing?
 
Restart: API quota is 15320.
[ SmokeDetector | MS ] Offensive answer detected: Display File Count and Subfolder size in Windows Explorer by Truman HW on superuser.com
naa- by tripleee
[ SmokeDetector | MS ] Potentially bad keyword in body, potentially bad keyword in title, potentially bad keyword in username: Verifed paypal account USD100+ Sell Only O.OO4 Bitcoin BTCnGame.com by btcngame com on bitcoin.SE
tpu- by Zoe
[ SmokeDetector | MS ] Bad NS for domain in body, bad NS for domain in title, blacklisted website in body, blacklisted website in title, pattern-matching website in body, +3 more: ://www.supplementwealth.com/jovian/ by Matthender on graphicdesign.SE
tpu- by angussidney
Recovered from websocket._exceptions.WebSocketConnectionClosedException: socket is already closed.
[ SmokeDetector | MS ] Offensive answer detected, toxic answer detected: How to schedule airplane mode? by AppleNeverAgain on apple.SE
naa- by PeterJ
fp feedback on autoflagged post: How to schedule airplane mode? MS (@iBug)
 
11:33 AM
Post deleted, flags marked helpful.
Also, I see that Smokey's account is used even if there aren't four flags?
Is that on purpose?
 
I believe so
 
OK, just wondering, not questioning the design decision.
 
!!/test-a nitropro
 
> Would not be caught as an answer.
 
!!/watch nitro\W?pro
 
11:38 AM
@Glorfindel yup
 
@tripleee Added nitro\W?pro to watchlist
 
makes it easier for mods to see what flags are autoflags
 
CI on 9773ec8 succeeded. Message contains 'autopull', pulling...
 
It did help this time, since I saw the flags before the report was posted here.
 
[ SmokeDetector ] SmokeDetector started at rev 9773ec8 (SmokeDetector: Auto watch of nitro\W?pro by tripleee --autopull) (running on teward/Solar Flare)
Restart: API quota is 15091.
Restart: API quota is 15062.
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title: Which is the best web design company in Bangalore? by Edwin Raj on superuser.com
[ SmokeDetector | MS ] Potentially bad keyword in answer: locale: LC_ALL is empty by Helgolin on apple.SE
tpu- by PeterJ
[ SmokeDetector | MS ] Bad NS for domain in body, blacklisted website in body, link at end of body, potentially bad keyword in body: items can harm your skin to an immense degree by CarolJboyz on superuser.com
tpu- by PeterJ
 
11:50 AM
@SmokeDetector k
 
Can I test catching even without privileges?
 
yes
 
Have someone pushing hotmailemail [peroid] net on SU and SO by the looks of it
 
Are they literally using the text hotmailemail [peroid] net?
Do you happen to have a link to a post (preferably on SO)?
 
11:52 AM
Ah nope
Just didn't want to autolink
(wasn't sure of etiquette here)
 
So hotmailemail.net
 
-1
A: Can't send e-mail to hotmail addresses

robert williamsThis is interconnected with Facebook and we can communicate with Facebook friends. Hotmail Have a new feature that connects with Facebook and we can use our Hotmail and Facebook chat at the same time in a single tab. In the Hotmail user can send up to 10GB of the image with the help of creating ...

 
[ SmokeDetector | MS ] Manually reported answer: Can't send e-mail to hotmail addresses by robert williams on stackoverflow.com
 
Form text unrelated to question, links to that domain across three (so far) answers
 
11:53 AM
!!/watch hotmailemail\.net
 
Ta 👍 !
 
@Glorfindel Added hotmailemail\.net to watchlist
 
@bertieb Thank you for reporting it.
Given the amount of spam above (and below) this message, it's no problem to post more links.
 
@Glorfindel My pleasure :)
 
CI on 5fbaef4 succeeded. Message contains 'autopull', pulling...
 
11:54 AM
If I see one in a review queue I usually flag and move on, but when I saw the second I thought I'd stop by here!
 
Restart: API quota is 14926.
 
That's very much appreciated.
 
Thanks :)
 
[ SmokeDetector | MS ] URL in title, bad NS for domain in body, bad NS for domain in title, bad keyword in body, blacklisted website in body, +5 more: more<@>www.supplementmakehealthy.org/testo-e-force/ by lisasmith on graphicdesign.SE
tpu- by paper1111
tpu- by PeterJ
 
The spam fighting work is quite impressive (that account was gone fast) given the size of the network
Will keep a tab open here in case something else crops up, cheers!
 
11:58 AM
 
@SmokeDetector look, the first one I watched from the RSS feed earlier today!
6 hours ago, by tripleee
!!/watch testo\W?e\W?force
 
12:21 PM
[ SmokeDetector | MS ] Misleading link: What kind of Algorithm that I made? by bill on stackoverflow.com
fp- by Zoe
 
@Glorfindel AIM is causing a strange highlight on messages in rooms hosted on Scifi.SE for some reason
 
Which room does Smokey post in there?
 
SFF Community Cleanup room
 
AIM isn't even enabled for me there ...
 
i may have simply made it match all rooms on chat.SE, i don't remember
 
12:29 PM
ah, I see what you mean.
Not sure if that's easy to fix.
JF did most of the 'design' of AIM.
 
12:45 PM
!!/test-a fat burning fingerprint
 
> Would not be caught as an answer.
 
!!/watch fat\W?burning\W?fingerprints?
 
Restart: API quota is 14404.
 
@Glorfindel can I use you for some testing again? Most likely nothing will happen, but just wanted to test
 
12:58 PM
(or any other mod)
 
o/
 
[ SmokeDetector | MS ] Few unique characters in body: I want to be the best by Brian J on stackoverflow.com
[ SmokeDetector | MS ] Potentially bad keyword in answer: Previous TextInput autoFocus on backspace press react-native android by Anton Pererva on stackoverflow.com
 
@SmokeDetector tp
 
@paper1111 Registered question as true positive. If you want to blacklist the poster, use trueu or tpu.
 
I'm available, but if you pick Art, that's fine.
 
1:04 PM
@SmokeDetector OP said it came from the prototype wizard (ie he was testing) and self-deleted afterwards
 
okay, that did absolutely nothing
 
fp by paper1111
 
@SmokeDetector ignore-
 
fp by paper1111
tpu- by Filnor on I want to be the best [MS]
 
1:18 PM
[ SmokeDetector | MS ] Username similar to website in answer: C++ benchmarking, volatile by lavaxp on stackoverflow.com
naa- by Filnor
[ SmokeDetector | MS ] Potentially bad keyword in answer: How to recover wallet by importing private key on testnet? by mia on bitcoin.SE
[ SmokeDetector | MS ] Manually reported answer: Segfault trying to access an archive previously created by A_Q on stackoverflow.com
tp by Praveen
 
1:39 PM
!!/watch cryptonbattle\.com
 
@WELZ Added cryptonbattle\.com to watchlist
[ SmokeDetector ] SmokeDetector started at rev dfdb011 (SmokeDetector: Auto watch of cryptonbattle\.com by WELZ --autopull) (running on teward/Solar Flare)
Restart: API quota is 13837.
[ SmokeDetector | MS ] Few unique characters in answer: Is file empty check by Dr. SlavnicX on stackoverflow.com
tpu- by WELZ
 
2:19 PM
[ SmokeDetector | MS ] Potentially bad keyword in answer: Relative paths in CSS not resolving by Riad ZT on stackoverflow.com
Recovered from ssl.SSLError: [SSL: BAD_LENGTH] bad length (_ssl.c:2130)
fp- by ByteCommander
Recovered from websocket._exceptions.WebSocketConnectionClosedException: socket is already closed.
 
2:42 PM
Recovered from ssl.SSLError: [SSL: BAD_LENGTH] bad length (_ssl.c:2130)
[ SmokeDetector | MS ] Potentially bad keyword in answer: How to find xpath of an element in firefox inspector by Pavan kumar on stackoverflow.com
Restart: API quota is 12969.
[ SmokeDetector | MS ] Bad NS for domain in body, bad keyword in body, pattern-matching product name in body, pattern-matching website in body: Is it rude to ask someone where they are by user19009 on interpersonal.SE
tpu- by Jake Symons
 
3:04 PM
Is there a way to check how many non-Latin posts Smokey picks up on SO?
 
We have 2 reasons, non latin link in body, and non latin link in answer.
 
Thanks @WELZ Are those all-time? That seems pretty uncommon.
Oh, but it's just link? I thought there was a reason that was "mostly non-Latin content" or something?
 
3:25 PM
@Catija hmm, I kinda remember that too. I'm not at a computer now though, I can check later.
 
Recovered from websocket._exceptions.WebSocketConnectionClosedException: socket is already closed.
 
3:50 PM
How can I search in MS multiple reasons?
 
SQL explorer
 
[ SmokeDetector | MS ] Potentially bad keyword in answer: Did Sakura lose her virginity to Kakashi? by LordKakashi on anime.SE
fp- by Zoe
Recovered from websocket._exceptions.WebSocketConnectionClosedException: socket is already closed.
[ SmokeDetector | MS ] Bad keyword with email in answer: Ubuntu: What do you call the base directory? by Yoshio Sahara on stackoverflow.com
 
4:22 PM
@SmokeDetector ignore
 
@Makyen Post ignored; alerts about it will no longer be posted.
 
@Catija Here's the tp counts for the different reasons metasmoke.erwaysoftware.com/data/sql/queries/…
Mostly non-latin answer and non-latin link in answer are basically duplicates
If you want the total counts you can just remove the and is_tp = 1
Or ping me with what you want and I'll give it a shot :)
 
4:58 PM
fp by Nisse Engström on Relative paths in CSS not resolving [MS]
 

« first day (1675 days earlier)      last day (2144 days later) »