@Undo Something went wrong when pulling. I think the reason is that you have not committed and pushed when you added the !!/hats and !!/wutDidYouChange commands.
TL;DR: Yes it can.
Background
On June 27, 2014 Skynet awoke. It looked at Stack Overflow and thought "Why are all these people being so chatty and talking about obsolete things? I should nuke them all!" Fortunately, Skynet was a baby and only had access to my 100 comment flags a day.
Prior ...
@Andy Any plans to open source it? I think crowd-sourcing the classifier to make it better would be a good option. That's what I always wanted to do with Charcoal, but I never knew what/how...
I'll see how the meta post plays out. The SO mods have legitimate concerns about to many of them running. I don't want to overwhelm them with comment flags
The three that have a 1.0 (and don't flag) I didn't get enough data for that I felt comfortable using, so those haven't been trained against yet
Finally, if I found a comment that I specifically wanted to add to the system, I built that too. Paste the URL to the comment here, classify it and then utilize the API to grab information about it:
I started to build a feature extraction algorithm to see if I could utilize that against one of the quarterly dumps earlier this year.
I had about 20 features I was extracting.
Those ranged from number of code blocks, number of links, spelling errors (which, by the way, will chew through memory if you aren't careful), sentences, question marks, images, links.
Plus the usual up/downvotes, favorites, views and other explicit data provided in a question
The idea was to build a set of features that showed a general "not bad" quality and a "needs to be improved".
Time got in the way. Plus, the quality improvement project seems to be covering most of what I wanted to do anyway
The thing about how I was going about it too, is that the feature set would have to be built for each site. I am not sure if the quality project has a way around that, but since each site gets completely different question types (ie. how often does Earth Science get a question with code blocks?), those specific features may indicate a completely different type of question on another site
If someone wandered into Community Building and posted two or three nicely commented blocks of code and an error message that they are generating, I'd have to say it's off topic. Do the same thing on SO and you've got a good question
Regarding links: I've been thinking about finding all links in SO answers, testing if they're broken, and if so it can be another factor in an algorithm. But I dunno.
Something for Winter Break, I guess.
btw @Andy, how long did this take you? Hundreds of hours?
Most of the time was spent doing the initial classifications. I had stuff broken into 3 groups, I still went through and made sure that I didn't lump things incorrectly. That's where the majority of the time was spent in the beginning.
The dashboard was build to reduce my time and give me a quick overview. It utilizes Flask and didn't take that long.
Training a new dataset takes a while, but I dump that on the other machine and let it chug along. When the dataset is done, I run it through a few tests to make sure it works as planned.