Low Quality Posts HQ

12:16 AM

Spam Flag Request for drupal.stackexchange.com/questions/135603/… by Jan Dvorak from Tavern on the Meta

12:26 AM

Close Vote Request for stackoverflow.com/questions/26694616/… by cVplZ from Tavern on the Meta Reason: and naa

6 hours later…

6:36 AM

Spam Flag Request for superuser.com/questions/698900/… by Jan Dvorak from Tavern on the Meta

1 hour later…

7:40 AM

Close Vote Request for stackoverflow.com/questions/623607/dynamic-sql-java-library by cVplZ from Tavern on the Meta

Close Vote Request for stackoverflow.com/questions/3871449/… by cVplZ from Tavern on the Meta

Close Vote Request for stackoverflow.com/questions/2770619/… by cVplZ from Tavern on the Meta

Close Vote Request for stackoverflow.com/questions/301246/… by cVplZ from Tavern on the Meta

1 hour later…

9:10 AM

Close Vote Request for stackoverflow.com/questions/26697502/… by TGMCians from Tavern on the Meta

9:48 AM

Close Vote Request for stackoverflow.com/questions/26697188/… by Sam from Tavern on the Meta

Close Vote Request for stackoverflow.com/questions/26696333/… by Sam from Tavern on the Meta

10:39 AM

Yay, finally finished moving all the terms to json.

10:54 AM

Flag Request -> Very Low Quality for bitcoin.stackexchange.com/questions/32281/is-this-true-100x-btc by Jan Dvorak from Tavern on the Meta

11:07 AM

Close Vote Request for stackoverflow.com/q/26698587/871050 by Second Rikudo from PHP

Close Vote Request for stackoverflow.com/q/26318421/2982225 by user245167 from Tavern on the Meta

11:33 AM

Darn, the scores are more messed up than I originally thought... (due to culture issues when Pham was running on Pat's machine)

11:54 AM

Close Vote Request for stackoverflow.com/questions/26575534/… by hakre from PHP

@Sam Can you just reset the scores?

Yes, it's possible.

But, I don't want to wipe out over a month's data.

I've wrote some code to auto-check and correct the scores, so hopefully that should be enough. If not, then I'll reset the scores.

12:35 PM

Close Vote Request for stackoverflow.com/questions/26699319/… by PeeHaa from PHP

1:24 PM

Ugh, I'll have to reset the scores. At least there'll be fresh data for the new stat commands.

Unihedron

1:45 PM

@Sam Ok.

2:20 PM

There, term scores all reset :/

Unihedron

Here goes hard work calibrating it..

At least we'll have a more effective/efficient system to play with ;)

Unihedron

3:18 PM

Close Vote Request for stackoverflow.com/questions/26700839/… by PeeHaa from HTML / CSS / WebDesign

Close Vote Request for stackoverflow.com/questions/26700839/… by PeeHaa from PHP

Close Vote Request for stackoverflow.com/questions/26700637/… by hichris123 from Tavern on the Meta

3:38 PM

Close Vote Request for stackoverflow.com/questions/26700973/… by PeeHaa from HTML / CSS / WebDesign

Close Vote Request for stackoverflow.com/questions/26701025/… by PeeHaa from PHP

Close Vote Request for stackoverflow.com/questions/26692927/… by hakre from PHP

Close Vote Request for stackoverflow.com/questions/26701097/… by hichris123 from Tavern on the Meta

Close Vote Request for stackoverflow.com/questions/26700828/… by Zach Saucier from HTML / CSS / WebDesign

4:30 PM

Close Vote Request for stackoverflow.com/questions/26418387/… by hakre from PHP

4:45 PM

Dev Progress Update: All terms have been reset to default scores (and are now stored as Json). The new stats command should be completed tonight (if not tomorrow).

5:22 PM

Spam Flag Request for meta.stackexchange.com/a/242948/213575 by Braiam from Tavern on the Meta

5:33 PM

Close Vote Request for stackoverflow.com/q/26702241/2301450 by vaultah from Python

5:48 PM

Close Vote Request for stackoverflow.com/questions/26700546/… by Zach Saucier from HTML / CSS / WebDesign

1 hour later…

6:53 PM

Close Vote Request for stackoverflow.com/questions/26702142/… by Benjamin Gruenbaum from JavaScript Reason: dupe

7:23 PM

Close Vote Request for stackoverflow.com/questions/26700983/… by tereško from PHP

@Jan Before I commit the new command, I just want to be doubly-triply sure that these algo's are correct...

Term Sensitivity = Term TP Count / Total Global TP Count

Term Specificity = Term FP Count / (Total Checked Posts Count - Total Global TP Count)

no scraping Zephyr?

Not at the moment. I'll add that functionality later.

your specificity formula doesn't seem to account for ignored reports

unless by "total checked posts count" you only mean those who were handled

Erm, I thought we agreed we'd leave out ignored posts and just have an inflated result.

7:32 PM

in which case this only applies if the set of handled posts were a representative sample of the whole population - which isn't the case.

You don't want to inflate the result. The idea was to distribute ignored posts between TPs and FPs

Ok...

Term Specificity = Term FP Count / ((Total Checked - Total Ignored) - Total TP Count)

Oct 30 at 18:19, by 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

to elaborate on #1: estimated_TP = (handled_TP/handled) * reported

@Sam so... nope to your formula

7:53 PM

Oh, I see. So: Specificity = Term FP / ((Total TP Count / Total Checked) * Total Reported)

nope; specificity = ((all fp - term fp) * term handled / term reported) / (all posts - all tp * all handled / all reported)

I assume handled = non-ignored.

yep