Print

[ethercat]

Rarefied Technologies Announces Public Beta of StupidFilter Language Processing Software

ALBUQUERQUE—July 22, 2008—Rarefied Technologies today announced the beta release of its highly-anticipated StupidFilter software. The StupidFilter aims to accurately identify deviation from standard written English, giving web application developers the tools they need to automate the moderation of community-driven sites in ways previously impossible.

Available now, this first public beta release of the StupidFilter supports classification—identification of text as either standard or substandard. Upcoming releases will be able to identify the precise degree of deviation from standard English using a technique called textual regression.

There's more at http://stupidfilter.org/main/index.php?n=Main.News
Here's a demo: http://stupidfilter.org/demo.php

[Lorelei]

I am supposedly a mod at the StupidFilter site, but the stupidity eventually wore me down and depressed me, and I have not contributed in a couple of months.

[Ultrapoet]

I am supposedly a mod at the StupidFilter site, but the stupidity eventually wore me down and depressed me, and I have not contributed in a couple of months.

Stupidity on the part of the people running it?  Or the stupidity you were testing to filter out?

[Lorelei]

Stupidity I was filtering. When I was modding, you were given 5-6 snippets of Stupid per page to rate on a scale from 1-5. Since almost all Stupid samples were 5s, I failed to see what sitting at my computer hitting "5, 5, 5, 5, 5, 5" for an hour every day really contributed to the project. Also, you weren't supposed to judge the stupidity of the content, per se, just the format / grammar / spelling / non-standardness of expression. Alas, a lot of the content was also "5, 5, 5, 5, 5, 5," which was depressing. The rating is done by double-blind group voting; you don't know what others voted nor do they know how you voted until after your vote is submitted. I noted that my standards tended to be a bit more harsh than other mods, and yet I couldn't let Stupid pass without a 5 it deserved just to be a "nicer" mod. (FWIW, the Stupids posting the Stupid were unaware their words were being rated.)

The people running it were fine; in fact, we barely interacted. If you did your job well, then all was well, and there was no need for aggromoderation to stamp out the merest whiff of variations in style between moderators. Unless someone was really being disruptive or deliberately trying to skew the data results, the folks running it were hands-off uber-moderators, choosing not to kick contributors in the teeth for slight deviations in style or opinion. Also, for the most part, Stupid is Stupid, and there were few people who did not agree on how certain Stupid samples should be rated.

I stepped down because I noted Stupidity slipping by other mods that was REALLY Stupid. Supposedly mods were (at least originally) chosen for their ability to judge non-standard communication and for their editing / teaching / writing backgrounds. It is possible that, in an effort to get more participants in the beginning, that SF let some non-standard mods into the pen along with the true Grammar Nazis.

[Ultrapoet]

Yeah, I could see how that would get depressing after a while.  I still think it sounds like a pretty nifty idea and I'll probably pounce on the WordPress plugin the moment it's made available.

[SocialTransparency]

 I went to the beta website, typed in on running sentences and mispelled words. Results came back as 'Stupid". Thing is the message I typed in was me being normal. Go figure! I truely am "With stupid".

[ethercat]

One problem is that just because someone doesn't type or speak English properly, or any number of other reasons their written text might not be up to snuff, doesn't mean they are stupid, it just means they can't communicate well with the written word.  Still, as a filter for comments sections, it will probably do pretty well, since on blogs, comment sections, etc., all there is is the written word.

It's similar to the Bayesian filters (called SpamProbe) I use for email, where I train the software what is spam and what isn't, then the words in the email are scored, and a "judgement" of whether it is spam or not is made.  It's usually pretty accurate, but sometimes it does miss.  I send the false positives and negatives back to retrain the software, using a script I called SPtrain (and I laugh everytime I run it).

I think the StupidFilter could still use some work.  I put OT3 through it, and it came back with "Text is not likely to be stupid."      Unfortunately, while there is a link for "If you get something classified as stupid incorrectly, please submit false positives here," there is no similar link for false negatives.   

[Lorelei]

It is not measuring the stupidity level of th con tent, just the mechanics of the written sample and how it measures up to standard English grammar. ESL folks are going to thwart the stupid filter, because English is not their primary language. It is not intended to be used for ESL, or to judge whether or not a particular piece of writing is patent nonsense.

I believe SF gets its samples mostly from YouTube comments, where the Stupid runs high, and you can't easily edit yourself if you make a typo.

[Ultrapoet]

I know ESL speakers who write more precisely in English than native speakers because they've been STUDYING it.

And I'm not the least bit surprised that the cesspits of YouTube comments was such a rich source of testable stupid.

[Lorelei]

You are correct: most ESL speakers I know personally are far more articulate when communicating via the written word than are the vast majority of presumably native English speakers I run across on the Internet.

[seataka]

  Would it be possible to run some of Hubbard's gibberish through that filter?

[SocialTransparency]

  Would it be possible to run some of Hubbard's gibberish through that filter?

You would need to figure out how to get your computer, via some strange form of osmoses, to upload 2 tin cans. Impress me!