Low Quality Posts HQ

Jan Dvorak

6:06 PM

eh... just distribute ignored reports with the same ratio as handled reports

then specificity = inflated fp / (total - known bad)

where known bad is (inflated (tp from any keyword)) + zephyr

that should work...

Sam

Ok, so we're leaving out ignored posts from our calcs completely?

Jan Dvorak

we're distributing ignored posts between TP and FP in the same proportion as real TP's/FP's

Sam

I see; alright, I'll get to work ;)

Jan Dvorak

#1: distribute ignored positives proportionally between TP and FP
#2: estimate that bad posts = those reported by Zephyr or TPs of other keywords
#3: specificity = TN / bad; sensitivity = TP / good

@sam You could also report %ignored separately

Sam

6:25 PM

I'll probably just append it the the end of stats results.

Jan Dvorak

(I've just realised that % ignored is keyword-specific)

Sam

Meh, it's easy to calc a global ignored metric.

Jan Dvorak

per-keyword is more interesting

I'm sure .{0.150} will get a lot of ignores ;-)

Sam

I'll add both ;)

Now, any thoughts for the command's name? I was thinking of >>stats (for global stats) and simply append the extra data to the why command (term specific), but I'd like to keep this a democratic process ;)

Jan Dvorak

I'm fine with anything

Sam

6:29 PM

Ok.

Jan Dvorak

could some part of it be strapped on to why?

Sam

Yeah, I've just edited my message.

Jan Dvorak

I'd try just strapping everything keyword-related to why

Sam

Even global info?

Jan Dvorak

not that

Sam

6:33 PM

Right, so sensitivity + specificity + ignored % for terms.

Jan Dvorak

yep

Sam

kk

Jan Dvorak

no more interest in accuracy then?

I'm fine with that

Sam

Hmm

I'll implement sens/spec/igno first.

Jan Dvorak

The problem of accuracy is that it fluctuates with bad/total

Sam

6:37 PM

True, I'll leave it out in that case. The other stats will be enough.

Jan Dvorak

more spam = more positives are true

Zephyr

6:54 PM

Close Vote Request for stackoverflow.com/questions/3146901/… by ProgramFOX from The SO Tavern (Free Snacks)

Jan Dvorak

@Sam note that if there are only few handled reports, the specificity and sensitivity readings will be very inaccurate; don't forget to leave them out if there are only few reports. (edge case: no reports handled means you don't know how to split the ignored ones)

Sam

Ok, shall we say a min of 5 reports then?

Jan Dvorak

should do

Sam

kk (I can always alter the limit if needs be)

Jan Dvorak

5 heads or 5 tails are sufficient to prove with 95% confidence that a coin is not fair.

Zephyr

7:01 PM

Close Vote Request for stackoverflow.com/questions/533231/… by ProgramFOX from The SO Tavern (Free Snacks) Reason: tool

Close Vote Request for stackoverflow.com/questions/198400/… by ProgramFOX from The SO Tavern (Free Snacks)

Sam

Lets see, I'll now need to keep a record of:

Total "checked" posts count,
Total TP count,
Each term's TP/FP/ignored/reported count.

Jan Dvorak

reported = tp + fp + ignored

why do you want total tp count?

Sam

57 mins ago, by Jan Dvorak

where known bad is (inflated (tp from any keyword)) + zephyr

Zephyr

Spam Flag Request for drupal.stackexchange.com/questions/135405/… by Jan Dvorak from Tavern on the Meta

Jan Dvorak

true

Sam

7:10 PM

It would be easier to just keep a global record, rather than counting every term's TP count.

Jan Dvorak

>>add-b-qb-spam chaussure

Sam

Lemme start Pham...

Jan Dvorak

oh.

Sam

When Pat left I didn't start him back up.

(due to the same old internet issues)

But, hopefully I should be getting it fixed this sunday :D

Pham

Phamhilator™ started.

Jan Dvorak

7:12 PM

>>add-b-qb-spam chaussure

Pham

Low Quality Q (62.9%): Cordova build, get an error, by SG87, on apple.stackexchange.com.

Low Quality Q (36.4%): What are non-heritable changes to genomes?, by Dissenter, on biology.stackexchange.com.

Sam

It takes a few secs for the command listener to "warm-up".

It should be fine now, though.

Pham

Low Quality Q (39%): What stops a developer from accessing credit card details and other secret data from a company, by Ayesh K, on security.stackexchange.com.

Low Quality Q (45%): Need help making jQuery Transit transition repeat, by John, on codereview.stackexchange.com.

Jan Dvorak

@Pham tpa

how is that too soon?

Pham

Low Quality Q (18.6%): confidence intervals and type II errors, by noname, on stats.stackexchange.com.

Sam

7:16 PM

sighs I'll check if pat changed the botusernam again...

>>infp

Pham

Command not recognised.

Sam

>>info

Pham

Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 941. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 749. Uptime: 00:04:03.4499245.

Sam

>>status

Pham

Low Quality Q (100%): I am habitual of listening this from you ? “Is this statement correct ”, by Utkarsha Tiwari, on english.stackexchange.com.

Sam

7:16 PM

>>help

Pham

See here for a full list of commands.

Sam

>>status

Pham

Current status: Beta.

Sam

@Pham fp

Yep, looks like a username issue, again.

Pham

Low Quality A (21.5%): spotlight - activity monitor - search for ..., by Wee Malcky, on apple.stackexchange.com.

Jan Dvorak

7:17 PM

the bot should autodetect their username

Sam

I was going to add that functionality, but got sidetracked :/

>>info

Pham

Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 941. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 751. Uptime: 00:06:18.2446344.

Phamhilator™ stopped.

Phamhilator™ started (debug mode).

Sam

>>pause

Pham

Low Quality A (21.5%): Try this killall htop, by vembutech, on askubuntu.com.

Bad Tag Used: [web] | Record audio on multiple tracks simultaneously from multiple devices, by Antonio Paredes Picon, on stackoverflow.com.

Bad Tag Used: [size] | how to crop image with plugin jquery, by Zombie, on stackoverflow.com.

Sam

>>pause

Pham

7:23 PM

Phamhilator™ paused.

Sam

^ just to reduce the chances of a collision.

>>info

Pham

Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 941. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 754. Uptime: 00:01:22.5367209.

Sam

Now try.

>>add-b-qt-lq test

Pham

@Sam Blacklist term added.

Sam

>>del-b-qt-lq test

Pham

7:26 PM

@Sam Blacklist term removed.

Sam

@JanDvorak Wouldn't it be a good idea to add case insensitivity?

Jan Dvorak

noted

please do

Sam

The term hasn't been added yet.

Jan Dvorak

oh.

Sam

>>add-b-qb-spam (?i)chaussure

Pham

7:31 PM

@Sam Blacklist term added.

Jan Dvorak

>>add-b-qb-spam (?i)chaussure

Pham

Blacklist term already exists.

Sam

Any more terms before I shut Pham down?

Jan Dvorak

why not leave it running overnight?

Sam

'cause I have to shutdown my lappy when I go to sleep.

Jan Dvorak

7:33 PM

go ahead then

Sam

Kk. I'll also add some more debugging capabilities to Pham (to see why exactly he keeps crashing on your pc).

>>kill-it-with-no-regrets-for-sure

Pham

Killing...

Command listener thread has died...

Warning: 10 seconds have past since the kill command was issued and post catcher thread is still alive. Now forcing thread death...

Zephyr

8:32 PM

Close Vote Request for stackoverflow.com/questions/26662671/… by Alex Thornton from Python

Sam

@Jan Have a few mins for some cooperative debugging?

Jan Dvorak

yep

Sam

Great, can you pull the latest version of Pham and try running him pls?

You should get a more descriptive error message this time.

Jan Dvorak

pulling

launching

Sam

sits in suspense

Jan Dvorak

8:39 PM

DirectoryNotFound

Sam

Ok, what's the error it copied to your clipboard?

Jan Dvorak

System.TypeInitializationException: Inicializační metoda typu Phamhilator.GlobalInfo vyvolala výjimku. ---> System.IO.DirectoryNotFoundException: Část cesty C:\Users\Honnza\Documents\GitHub\phamihilator\Question\Body\Black Filter Terms\Spam Terms.txt nebyla nalezena.
   at System.IO.__Error.WinIOError(Int32 errorCode, String maybeFullPath)
   at System.IO.FileStream.Init(String path, FileMode mode, FileAccess access, Int32 rights, Boolean useRights, FileShare share, Int32 bufferSize, FileOptions options, SECURITY_ATTRIBUTES secAttrs, String msgPath, Boolean bFromProxy, Boolean useLongPath, …

oh. It overrode my clipboard contents... That's unexpected

oh, and don't you love localised error messages?

Sam

I had to, that message was caught by the global exception handler, so...

Now I've just gotta translate the ex...

Jan Dvorak

path part ...\spam terms.txt not found

Sam

Yeah, but I need the entire ex (e.g. the trace).

Thank goodness for google translate :D

Jan Dvorak

8:44 PM

This is all I've got :-/

Sam

np, it should be enough.

Does C:\Users\Honnza\Documents\GitHub\phamihilator\Question\Body\Black Filter Terms\Spam Terms.txt exist?

Or, how much of the path exists?

Jan Dvorak

I don't have GitHub\phamihilator\Question but I do have GitHub\phamihilator\Phamihilator\Question

phamihilator\Phamihilator is where the .exe lives

Sam

And what's in the upper pham dir?

Jan Dvorak

git files, licence, readme.md

and the project file

Sam

Right, same as me.

Jan Dvorak

8:49 PM

there is a Question directory but only Title inside

Sam

Did you pull from master or non-windows?

Jan Dvorak

phamihilator\Phamihilator\Question contains both \Body and \Title

Sam

Good, it should do. So was this non-windows or master?

Jan Dvorak

I didn't set a remote branch

Sam

It needs to be set to master for it to work.

Jan Dvorak

8:53 PM

huh... how do I check which remote branch is being tracked?

Sam

Please tell me you're using GitHub for windows (the GUI app).

Jan Dvorak

Yep

Sam

phew

Jan Dvorak

I'm using the git shell

Sam

Ok, well if you switch back to the gui.

Jan Dvorak

8:54 PM

hold on...

Sam

You should see which branch you're currently on above the commit history (the text is blue).

Jan Dvorak

the shell shows the local branch is master

Sam

Hmm. Have you just switched or something? Or was it on master the whole time?

Jan Dvorak

I've never changed branches. Could git reset do that?

Sam

Erm, dunno. But I wouldn't have thought it would change you to a non-default branch.

Ok, if you can open up...

DirectoryTools.cs

Jan Dvorak

8:59 PM

The gui app doesn't know about the repo since I've created it through the shell :-/

Sam

And tell me what you see on line 10.

You didn't clone?

Jan Dvorak

> private static readonly string root = Environment.CurrentDirectory;

I cloned through the shell

Sam

Ok.

Do you have an IDE for .Net stuff?

(I.e., VS or SharpDevelop)

Jan Dvorak

kinda; sharpDeveloper doesn't like the project and I haven't actually launched VSExpress yet

launching

Sam

Right, well I need to know what's in root when Pham's launched (just set a breakpoint on line 11).

Are you familiar with VS?

Jan Dvorak

9:04 PM

nope, but I've already used netbeans and eclipse for java

Sam

kk

Jan Dvorak

VS is preparing for the first use :-/

Sam

facepalm

That ~~could~~ does take a few mins :/

Jan Dvorak

This may or may not take a few minutes

I'll let you know once it's done launching

Sam

Grr, gtg for 5-10 mins (I guess that'll give VS time to get ready). I'll be right back...

kk.

Jan Dvorak

9:08 PM

why does VS have screaming menus?

Dropped the project file into VS. Now thrashing...

@Sam the debug command complains about build errors

huh... when I choose "debug using the last successful build", Pham starts up just fine...

Phamhilator™ started (debug mode).

Low Quality Q (77.1%): baby reflux problem, by anish, on parenting.stackexchange.com.

after having VS complain about the lack of symbols

Low Quality Q (100%): What does it mean “Break Ya Neck” and “Protect Ya Neck”, by Broly, on ell.stackexchange.com.

Low Quality Q (100%): What is the relation between multiple-regression and pearsons r?, by ivanski dobrovski, on stats.stackexchange.com.

@Sam this smell of some heisenbuggery

Low Quality Q (60%): Why does “Tap to click” not work in many games?, by SofaKng, on apple.stackexchange.com.

Low Quality A (21.5%): IDK I was wondering the same thing, by gracc, on apple.stackexchange.com.

Low Quality Q (62.9%): Error: Model in Pspice is undefined., by Millares, on electronics.stackexchange.com.

Bad Tag Used: [facebook] | Is there a way to get favorite music (or liked music) of the fans of a Facebook page?, by Joonas Virtanen, on stackoverflow.com.

Low Quality Q (60%): SQL Server decryption authentication not working with concatenation, by user3469285, on dba.stackexchange.com.

Low Quality A (21.5%): where i can download startup-disc creator fo..., by Annamari Ketola, on askubuntu.com.

Low Quality Q (100%): Adafruit gps breakout module help, by user3594666, on arduino.stackexchange.com.

Sam

9:33 PM

Back.

@JanDvorak lol, must be part of the new metro styles.

@JanDvorak What are the errors?

I only get a few warnings.

Jan Dvorak

Low Quality A (21.5%): Suck a cock. Hhhhhhhhhhhhhhhh hhhhhhhhhhhhhh..., by Ben, on apple.stackexchange.com.

Low Quality A (21.5%): For my QGIS 2.4 this is located in: W..., by Juan Rodriguez, on gis.stackexchange.com.

Sam

@jan Wait, have you changed GlobalInfo.BotUsername yet?

@JanDvorak tpa

Jan Dvorak

@Sam nope

Sam

You'll need to change it to Jan Dvorak until I can add auto username detection.

Jan Dvorak

It did announce to the tavern

Sam

9:36 PM

Nope, that was me.

Jan Dvorak

Low Quality Q (19.9%): Need help identifying a strange symbol in a wiring diagram, by hakonrk, on electronics.stackexchange.com.

Phamhilator™ stopped.

oh. I'll set up a second account first, though

Sam

Kk

Jan Dvorak

Will you run it now?

Sam

I need to, so I can add "cock" to the filter.

Pham

Phamhilator™ started.

Jan Dvorak

9:39 PM

place your bets. How many FP's on bio.se?

Sam

>>add-b-a-off (?i)\bcock\b

Pham

@Sam Blacklist term added.

Sam

I've added word boundaries, so we shouldn't get too many ;)

Pham

Phamhilator™ stopped.

Jan Dvorak

night

Sam

9:41 PM

Night!

Oh, btw, don't forget to sync the filters when you finish using Pham.

Jan Dvorak

There haven't been any changes so far

should I commit "latest x titles"?

Sam

When you're up and running properly, yes.

But for tonight, don't worry about it.

Jan Dvorak

for now I'll just promise to reset my changes

Sam

kk

Jan Dvorak

what do you use to raise socks, btw? An anonymous window?

Sam

9:45 PM

Well, I actually used another browser.

Sam

9:55 PM

I'll start working on the auto username detection...

Pham

10:28 PM

Phamhilator™ started (debug mode).

Sam

>>info

Pham

Owners: Sam, Unihedron, Patrick Hofman, Jan Dvorak & ProgramFOX. Total terms: 943. Accuracy threshold: 17.5%. Full scan enabled: True. Posts caught over last 7 days: 754. Uptime: 00:02:21.6421014.

Sam

@Pham fp

Yes! Auto username detection works!

I better commit...

Pham

Low Quality Q (100%): Vampire stake protection, by Liam Brown, on scifi.stackexchange.com.

Bad Tag Used: [facebook] | PFFacebookUtils logInWithPermissions fails on simulator, by Eddy, on stackoverflow.com.

Phamhilator™ stopped.

Zephyr

11:22 PM

Close Vote Request for stackoverflow.com/questions/58649/… by Cole Johnson from Tavern on the Meta

Transcript for