Charcoal HQ

iBug

2:00 PM

@ArtOfCode I think the template has some inaccuracy.

The FAQ says a small part of rudeness should be edited out, instead having the whole post flagged as R/A.

ArtOfCode

yeah

this isn't something we're going to do for every mishandled post

only if there's a persistent problem

iBug

The question is, is there any community that still don't know how to handle spam?

CalvT븃

There probably is, but they don't get much spam so isn't an issue

SmokeDetector

[ charcoal-se.github.io ] continuous-integration/travis-ci/push success on 4d374f4: The Travis CI build passed

[ 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: getsupplementhelp.com/tryvexin/ by rozella spore on astronomy.SE

tpu- by CalvT븃

[ SmokeDetector | MS ] Few unique characters in answer, no whitespace in answer, repeating characters in answer: I am trying to reference a merged cell on one worksheet and have it copy over into the next sheet by Poop on superuser.com (@iBug)

tpu- by iBug

iBug

And it's a mystery why the 3-flag mechanism still persists on ELU and Workplace.

WELZ

2:09 PM

Just want to point out, if brought down to 3 flags, it will be even quicker (so long as there are a few people online in chat here) 1 autoflag, and when you provide feedback through FIRE it will flag as well, so those posts will probably go in less time

SmokeDetector

[ SmokeDetector | MS ] Blacklisted website in body, link at end of body, link following arrow in body, pattern-matching website in body, repeated URL at end of long post: How should I get highest possible benefit? by Theite 19 on superuser.com (@iBug)

tpu- by iBug

[ SmokeDetector | MS ] Potentially bad keyword in answer: Why, in the US, is Japanese brutality ignored compared with the Nazi brutality in WW2? by a deleted user on history.SE

fp- by iBug

iBug

Community ♦ bump bug

Also, when the question or one of the answers is updated, the whole Q&A gets scanned again.

So say, a regex-ly suspicious answer is under a question where there's another frequently updated answer, it'd repeatedly generate a lot of FPs.

ArtOfCode

2:25 PM

just send an ignore feedback

WELZ

@SmokeDetector ignore

SmokeDetector

@WELZ Post ignored; alerts about it will no longer be posted.

David

where is the "!!/watch" and other chat commands documented?

WELZ

charcoal-se.org/smokey/Commands

iBug

@David !!/help

David

2:31 PM

@WELZ thanks

SmokeDetector

[ SmokeDetector | MS ] Link at end of answer: gdal_translate() function in R not working, but no error message by Sam on stackoverflow.com (@iBug)

fp- by iBug

David

!!/watch dogshunt.com

SmokeDetector

@David You don't have code privileges, but I've created PR#1557 for you.

David

@SmokeDetector oh thanks. (why can't I talk to a bot?)

CalvT븃

@David Please remember to escape the . - so dogshunt\.com

David

2:34 PM

right.

iBug

@SmokeDetector And drop the http://www. prefix as well.

CalvT븃

@David You can feedback (mark as spam or not), but not add to the blacklist without someone reviewing it

David

so just give it a post url.

CalvT븃

Or if Smokey reports a post, mark it with k (spam) or f (not spam)

!!/watch dogshunt\.com

SmokeDetector

@CalvT븃 Added dogshunt\.com to watchlist

David

2:37 PM

so say I had given it the Q url, what would it of done with it?

CalvT븃

@David Approved it

SmokeDetector

CI on 1fe8681 succeeded. Message contains 'autopull', pulling...

iBug

@David It will show it as "Manually reported question"

SmokeDetector

[ SmokeDetector ] SmokeDetector started at rev 1fe8681 (SmokeDetector: Auto watch of dogshunt\.com by CalvT븃 --autopull) (running on bwDraco/Linode)

Restart: API quota is 10068.

David

thanks all.

CalvT븃

2:39 PM

@David Sorry misunderstood. If you !!/report <question url> a post, then we would go and flag it

It would then go in our records, and we would potentially alter our spam catching system to catch more posts like it

David

@CalvT븃 so it does in some way train the bot. that is what I wanted to know, because being a mod I don't need the auto flag part.

CalvT븃

@David Well it doesn't exactly train the bot per se, but bringing it to our notice means we edit the filters to catch more like it

Say you'd reported the dogshunt question, we would of then watched the url (as you did), then any posts that include that url will now be automatically reported here

David

alright. I'll try to remember to !!/report post URLany spam in the future.

CalvT븃

That's awesome thanks

David

we do get some odd ones.

CalvT븃

2:44 PM

@David here's a list of what we've found for blender.se in the past metasmoke.erwaysoftware.com/…

iBug

Why "train" the bot? Aren't we supposed to "plane" the bot or "car" the bot?

CalvT븃

Only 48 posts, not a lot

@iBug train as in training

The verb, not the noun

iBug

(apparently joking)

ArtOfCode

@iBug (don't worry I got it :) )

CalvT븃

I've been away from England too long :/

iBug

2:46 PM

@CalvT븃 I even though you're an Argentinean.

CalvT븃

@iBug No - English born

SmokeDetector

CI on 5a6433c succeeded. Message contains 'autopull', pulling...

[ SmokeDetector ] SmokeDetector started at rev e344964 (ArtOfCode: Merge pull request #1558 from iBug/patch-2) (running on bwDraco/Linode)

Merged SmokeDetector #1553.

Restart: API quota is 9962.

Merged SmokeDetector #1551.

iBug

Feeling solo

SmokeDetector

CI on 7eadb9e succeeded. Message contains 'autopull', pulling...

[ SmokeDetector ] SmokeDetector started at rev 7eadb9e (metasmoke: Merge pull request #1551 from Charcoal-SE/auto-blacklist-1516760244.4658086 --autopull) (running on bwDraco/Linode)

Restart: API quota is 9936.

quartata

I'm kind of astounded by how few actively maintained stats/ml Ruby gems there are

ArtOfCode

3:14 PM

@quartata Ruby isn't really a stats/ML language

you're looking for Python

CaffeineAddiction

@ArtOfCode snarkyness asside ... what is Ruby good at?

ArtOfCode

Rails

CaffeineAddiction

what does it do better than python

ArtOfCode

I'll take Rails over Flask or Django any day

SmokeDetector

[ SmokeDetector | MS ] URL-only title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +3 more: perfecttips4health.com/glovella/ by uanuahmanxpa on askubuntu.com

JAD

3:22 PM

@ArtOfCode or R :)

SmokeDetector

[ SmokeDetector | MS ] URL in title, bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, +3 more: dietplanusa.com/power-testo-blast/ by Foultiam on graphicdesign.SE

tpu- by WELZ

tpu- by WELZ on http://perfecttips4health.com/glovella/ [MS]

quartata

3:53 PM

@ArtOfCode Of course not. But I'm not writing in anything else for something that Metasmoke would run

ArtOfCode

@quartata this sounds like scope creep

what are you planning?

quartata

18 hours ago, by John Dvorak

or perhaps go fancy, and pass the reason vector to a neural network to get a confidence rating

I decided to take this a little bit more seriously

just to try it

this is the best I found so far incidentally: github.com/tangledpath/ruby-fann

ArtOfCode

@quartata yeah, scope creep

metasmoke stores posts, it doesn't do in-depth analysis on them

quartata

This operates on the post's reasons

not the contents of the post itself

it's not really any more creepy than what we do for current autoflagging

ArtOfCode

same difference

quartata

3:56 PM

not really

ArtOfCode

Yeah

not something that metasmoke should be doing

make it part of Smokey or a separate bot

quartata

.....why?

what metasmoke is doing currently is a crude form of classification

how is making it a little less crude suddenly outside of its scope

ArtOfCode

It's adding some numbers from a database up, comparing them to other numbers in the database, and flagging if they work. It's almost not analysis at all.

In-depth analysis on machine learning lines is not something it was ever made to do.

(nor something that the server is likely to have capacity for)

quartata

I actually suspect it'd be faster than current autoflagging but

ArtOfCode

Faster than adding numbers up? Doubt it

quartata

4:00 PM

uhhh and several joins

ArtOfCode

@quartata which query, specifically?

quartata

calculating reason weight involves joining posts and posts_reasons right?

ArtOfCode

yep

that's about the fastest join we do anywhere in metasmoke :P

quartata

but it's still one

ArtOfCode

730 msec for all 100k posts

sub-10 msec for our standard 100 pagesize

quartata

4:03 PM

I just suspect it'll be tough to beat like maybe 3 matrix multiplications done in C is all

it does help that the reason weight is precalculated per week, I forgot about that

ArtOfCode

@quartata try it. You've got <10ms on 100 posts to beat.

quartata

that's just the DB query itself

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in link text in answer: Flight Search from anywhere to a specific destination by Smarty Jr on travel.SE

tpu- by CalvT븃

ArtOfCode

@quartata what's the equivalent operation?

quartata

looks like this fann benchmark is done on pretty bad hardware

20 nanoseconds per neuron would be in the microseconds range, not counting whatever overhead there is from the ruby binding

@ArtOfCode I'd tell you if I could find it... not sure where the reason weight pluck is actually done

at any rate the performance wasn't the point

the point was scope creep

which I have to disagree with

ArtOfCode

4:13 PM

cc @Undo

quartata

what metasmoke is doing now is still basically learning a distribution

the thing is that currently all reasons are independent

Undo

Can't read everything right now, but Standard Procedure applies: If we can make something that works better, objectively, than what we have, we can find a place to put it.

quartata

I was thinking of something probably more annoying before Dvorak just suggested a full NN, I figured it was at least worth entertaining

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, bad keyword in username, blacklisted website in body, link at end of body: Supreme Boostr - It will help you to reduce your muscle soreness and make your workout session longer by Supreme Boostr on arduino.SE

Undo

Remember that speed in autoflagging isn't a huge concern. That code only really runs on post creation, and those are relatively infrequent and delay-tolerable to the seconds range

SmokeDetector

4:20 PM

tpu- by CalvT븃

[ SmokeDetector | MS ] Link at end of answer: How to play a YouTube clip in background/minimised? by Du Bawshing on android.SE (@iBug)

tpu- by CalvT븃

quartata

well ruby-fann just segfaulted immediately on me, fun

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, pattern-matching website in body: Hydravella Eye Cream As you age by ansomaha on superuser.com (@iBug)

tpu- by CalvT븃

quartata

oh nevermind, it's just a little persnickety about the input type apparently...

ok so that works

at least it learned XOR

SmokeDetector

4:37 PM

[ SmokeDetector | MS ] Username similar to website in answer: Comparison of P2P Technology by Customer interations on stackoverflow.com (@iBug)

tpu- by CalvT븃

[ SmokeDetector | MS ] Link at end of answer: Выполнение php скрипта с планировщик задач на windows by Solutions IT on ru.stackoverflow.com

Henders

5:13 PM

TIL: h = 'hello',; type(h) outputs <class 'tuple'> in Python.

Raghav Dinesh

Hi SE!

I new to chat, hope mistakes will be forgiven.

am*

CalvT븃

@RaghavDinesh Hi

This is a chatroom dedicated to spam fighting

Raghav Dinesh

I have a question, Although I don't have any intension to break rules here but still I am afraid to ask questions. Becuase its really hard to earn reputation after a long day at work.

Hi @CalvT븃

CalvT븃

You don't earn reputation in chat

Raghav Dinesh

Yes

I agree but else where

How do I overcome fear of loosing reputation and ask questions

CalvT븃

5:19 PM

This isn't the only chatroom - find the chatroom for the site you want to ask a question on, then ask them

Raghav Dinesh

Answer as well as edit.

Oh great idea. One more question for you. Sorry I am taking your precious time.

If I ask questions here then I will hardly earn any reputation.

I am bothered about reputation because it unlocks new features at SE.

CalvT븃

We aren't part of any site, so there's no way you can earn reputation in Charcoal

Raghav Dinesh

Oh

Mithrandir

Yes. If you're seeking general advice, you're better off somewhere else I'm afraid.

Raghav Dinesh

I get it. Thanks

Mithrandir

5:22 PM

If you're looking for help with spam, then we're here.

Raghav Dinesh

Yes, thanks for being nice and for your time.

Thanks @Mithrandir @CalvT븃

SmokeDetector

[ SmokeDetector | MS ] Potentially bad keyword in title: Using Python to (sort of) Hack XML from a Federal API Call by Anders Engnell on stackoverflow.com (@iBug)

Jake Symons

I've been away for a couple of days on the chat, anything interesting happened?

SmokeDetector

fp- by doppelgreener

doppelgreener

@JakeSymons business as usual

yesterday, by Mithrandir

Every single page at http://charcoal-se.org/ should now be available at the Wayback Machine. If you find one that I missed, please add it ;)

Oh, actually:

Andy

5:35 PM

@Henders docs.python.org/3/tutorial/…

doppelgreener

2 days ago, by ArtOfCode

Hear ye hear ye, behold our new logo

2 days ago, by ArtOfCode

2

(but you might have been there for that)

Henders

@Andy Yep, an interesting surprise for me when I read about why it was happening :P

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in answer, bad keyword with email in answer, email in answer, pattern-matching email in answer: What sense of "hack" is involved in "five hacks for using coffee filters"? by Billijo Blewett on english.SE

fp by iBug on Выполнение php скрипта с планировщик задач на windows [MS]

[ SmokeDetector | MS ] Few unique characters in body, repeating characters in body: Python - how to fill a pygame screen from a different method by Harnell on stackoverflow.com (@iBug)

[ SmokeDetector | MS ] Few unique characters in body, repeating characters in body: SQL Query - Multiple different sums in a single column by Harnell on stackoverflow.com (@iBug)

tpu- by Jake Symons on What sense of "hack" is involved in "five hacks for using coffee filters"? [MS]

[ SmokeDetector | MS ] Few unique characters in body, repeating characters in body: Python syntax error on while loop while generating a random number by Harnell on stackoverflow.com (@iBug)

[ SmokeDetector | MS ] Few unique characters in body, repeating characters in body: What is the HTML equivalent to triple quotes in Python? by Harnell on stackoverflow.com (@iBug)

tp- by Henders on SQL Query - Multiple different sums in a single column [MS]

tp- by Henders on Python - how to fill a pygame screen from a different method [MS]

tp- by Henders on Python syntax error on while loop while generating a random number [MS]

doppelgreener

@SmokeDetector vandalism

SmokeDetector

tp- by Henders

@doppelgreener Registered question as true positive. If you want to blacklist the poster, use trueu or tpu.

SmokeDetector

6:01 PM

[ SmokeDetector | MS ] Link at end of answer: "Could not resolve host: download.docker.com" while installing docker CE by Rashmi on stackoverflow.com (@iBug)

[ SmokeDetector | MS ] No whitespace in answer: Android Play store additional content download location? by Babang Zahwwa Open on gaming.SE

tpu- by CalvT븃

[ SmokeDetector | MS ] Potentially bad keyword in body: Problema de conexion con webservice WSFEv1 - VB.NET by FrancoS on es.stackoverflow.com

fp- by CalvT븃

SmokeDetector

6:46 PM

[ SmokeDetector | MS ] Manually reported answer: Overloaded functions in python? by hEL on stackoverflow.com (@iBug)

tpu- by WELZ

[ SmokeDetector | MS ] Blacklisted user: Overloaded functions in python? by hEL on stackoverflow.com (@iBug)

[ SmokeDetector | MS ] Bad keyword with a link in answer, blacklisted user: Overloaded functions in python? by hEL on stackoverflow.com (@iBug)

tpu- by WELZ on Overloaded functions in python? [MS]

tpu- by WELZ

WELZ

seriously?!

Jake Symons

@SmokeDetector Theres one above as well, although that's probably gone now as well

When is he planning on taking over the website, after they have deafeated Smokey and the millions of other poeple

SmokeDetector

[ SmokeDetector | MS ] Mostly dots in answer: Java collision detection bug by Chiranjeevi Golu on stackoverflow.com (@iBug)

naa- by WELZ

WELZ

How come reporting as NAA (through FIRE) doesn't flag as NAA

Do we even have autoflagging for NAA?

Byte Commander

no

Smokey is a spam/abuse bot. If we catch some NAA as a side-effect it's okay, but that's not its purpose.

WELZ

7:01 PM

also the r/a will just flag as spam (which is technically incorrect)

Byte Commander

r/a will not flag at all, if you mean the button in FIRE

we have the approval of a community manager though that spam-flagging abusive posts is perfectly acceptable.

btw sorry if some messages might be duplicated, I'm on a train wifi with ~30kB/s bandwidth on average...

SmokeDetector

API quota rolled over with 6930 requests remaining. Current quota: 19999.

math: 654
stackoverflow: 525
superuser: 133
physics: 122
english: 93
serverfault: 83
codegolf: 70
askubuntu: 58
unix: 56
drupal: 44
apple: 41
interpersonal: 38
es.stackoverflow: 31
salesforce: 30
bitcoin: 30
rpg: 29
security: 28
academia: 28
travel: 26
electronics: 25
meta: 25
gis: 24
wordpress: 23
dba: 21
graphicdesign: 21
stats: 21
ethereum: 20
worldbuilding: 20
gaming: 20
sharepoint: 19
softwareengineering: 19
android: 18
money: 17
ru.stackoverflow: 17
diy: 16
islam: 15
mathematica: 15
history: 15

[ SmokeDetector | MS ] Blacklisted user: Save an Outlook Note without closing it by DukeSilversJazz on superuser.com (@iBug)

WELZ

7:20 PM

sd why

SmokeDetector

[:42471363]
Blacklisted user - blacklisted for http://superuser.com/a/1283950 (https://m.erwaysoftware.com/posts/by-url?url=//superuser.com/a/1283950) by https://chat.stackexchange.com/transcript/11540?m=42444036

fp- by WELZ

WELZ

!!/rmblu superuser.com/users/493349/dukesilversjazz

SmokeDetector

@WELZ User removed from blacklist (493349 on superuser.com).

SmokeDetector

7:34 PM

[ SmokeDetector | MS ] Bad keyword in body, blacklisted website in body, pattern-matching website in body: httpworldmusclekingcommusclescience by betcheyer45 on graphicdesign.SE

tpu- by Glorfindel

SmokeDetector

8:05 PM

[ SmokeDetector | MS ] Repeating characters in answer: Do you comment your code? by Kevin on stackoverflow.com (@iBug)

fp- by Glorfindel

[ SmokeDetector | MS ] Misleading link: typescript issue in importing the signature pad by dpdp on stackoverflow.com (@iBug)

fp- by Glorfindel

[ SmokeDetector | MS ] Bad pattern in URL answer: What to do if there is no gym or pokestop? by down_quark on gaming.SE

[ SmokeDetector | MS ] Bad keyword with email in body: I am looking for R studio Help for Econometrics homework - I will pay for it by Alf Alsudairy on stackoverflow.com (@iBug)

tpu- by WELZ

[ SmokeDetector | MS ] Few unique characters in answer, repeating characters in answer: When blowing air through a tube, why does it act differently if I press the tube against my mouth, or hold it an inch away? by mmm on physics.SE

tpu- by WELZ

fp- by WELZ on What to do if there is no gym or pokestop? [MS]

Henders

8:30 PM

@ArtOfCode How come every time I use Rails there's always some error on bundle install

ArtOfCode

shrug

you're probably missing a dependency, either for mysql or nokogiri

Henders

I kinda expected it to work on a fresh install of Ubuntu... :P Bleugh

ArtOfCode

nah, dependencies

Henders

Should I ever use sudo when installing bundles

ArtOfCode

I've never needed to

Henders

8:37 PM

ruby-dev was missing I think

WELZ

btw @ArtOfCode I got 5 flags today, I guess it was just coincidence (and that's about normal as well)

SmokeDetector

Restart: API quota is 19021.

Henders

9:18 PM

Hmm.. seems to just be a case of installing each thing it asks for when it fails...

SmokeDetector

[ SmokeDetector | MS ] Offensive answer detected, potentially bad keyword in answer: Macbook Air Aluminum Wearout by sex on apple.SE

ArtOfCode

@Henders pretty much

libmysqlclient-dev is a good one to have

Henders

Is MS rails 4 or 5?

SmokeDetector

[ SmokeDetector | MS ] Potentially bad keyword in answer: Can I buy a Bitcoin from one exchange and sell it on another? by Archer Carmona on bitcoin.SE

ArtOfCode

@Henders edge

so 5 point whatever the latest non-alpha is

Mithrandir

9:24 PM

...okay, a couple things about charcoal-se.org//ms.

1.) Why is there a double slash?

2.) Why is there no content?

3.) Why does that link to MS itself not linkify?

4.) Why does Smoke Detector have a space there?

ArtOfCode

1) because someone, dunno who, wrote the redirect from /ms wrong

2) it's a copy of the GH wiki, github.com/Charcoal-SE/metasmoke/wiki

3) see 2

4) see 3

Mithrandir

3.) It's linked on GH...

ArtOfCode

oh

it's a raw link

GH links raw links, website copies don't

Mithrandir

5.) Who wants to fit it?

ArtOfCode

you do \o/

Mithrandir

9:35 PM

*eyeroll*

I wouldn't know where to start.

CaffeineAddiction

you start with git clone git@github.com:Charcoal-SE/SmokeDetector.git

SmokeDetector

tpu by Mithrandir on Macbook Air Aluminum Wearout [MS]

tpu- by WELZ on Macbook Air Aluminum Wearout [MS]

tpu by Mithrandir

tpu- by WELZ

WELZ

!!/blacklist-website liviacoins\.com

SmokeDetector

@WELZ The !!/blacklist command has been deprecated. Please use !!/blacklist-website, !!/blacklist-username,!!/blacklist-keyword, or perhaps !!/watch-keyword. Remember to escape dots in URLs using \.

@WELZ You don't have code privileges, but I've created PR#1559 for you.

Merged SmokeDetector #1559.

[ SmokeDetector ] SmokeDetector started at rev af99430 (metasmoke: Merge pull request #1559 from Charcoal-SE/auto-blacklist-1516829956.2259824 --autopull) (running on bwDraco/Linode)

Restart: API quota is 18539.

ArtOfCode

9:51 PM

@Mithrandir edit the wiki

Mithrandir

That's not what I meant.

ArtOfCode

@Mithrandir oh. What did you mean?

Mithrandir

What would I fill it with? What content should be there?

SmokeDetector

[ SmokeDetector | MS ] Bad keyword in title, potentially bad keyword in body: Xavier vs Marquette Live Stream NCAA Men's College Basketball Game 24 Jan 2018 Online TV by NK Boishak on drupal.SE

tpu- by Mithrandir

[ charcoal-se.github.io ] continuous-integration/travis-ci/push success on 4d374f4: The Travis CI build passed

WELZ

@Mithrandir I fixed that

@ArtOfCode can you please put me on /people, so that next time they send swag, my greedy self will be satisfied.

ArtOfCode

10:02 PM

done

SmokeDetector

[ charcoal-se.github.io ] continuous-integration/travis-ci/push success on 219f6b6: The Travis CI build passed

[ SmokeDetector | MS ] Bad keyword in answer, bad keyword with email in answer: Where did Japan get their oil during WWII? by NefteGaz Agent on history.SE

tpu- by Mithrandir

WELZ

@ArtOfCode thanks

what do I write to include a keyword whether or not there's a space (in the middle of the keyword-s)?

Mithrandir

exam\W?ple

WELZ

Is there a list with all the (common) codes to use (and how/where) somewhere?

Mithrandir

@WELZ regexone.com

2

WELZ

10:15 PM

that was quite hard for me to figure out, wow

SmokeDetector

10:33 PM

[ SmokeDetector | MS ] Blacklisted website in answer: Where do I go looking for an editor? by KeyClavis on writers.SE

J F

@SmokeDetector I’m sorry, but World War II has finished. If you’d like to help supply Japan with oil during World War II, please use a time machine. However, your nation may be extremely unhappy with your support of the Japanese war effort.

Henders

11:03 PM