Here’s the link:
Is this feasible? It reads like an April Fools joke. Would it amount to anything more than a grammar filter? Could it detect irony and allow for humor? What are “Bayesian Rules”? Could it be improved to make an Ignorance Filter?
I’m a friend of the author, but I haven’t used his software. Bayesian rules are used for text filtering to find strings of characters that match a property. If a user highlights a string of leet-speak and flags it as “90% stupid” then the computer adds those strings to its corpus of “stupid” data. Eventually it has enough strings with enough scores to start inferring which strings are most likely to cause someone to say “this is really stupid.” “LOLOLOLOL” will probably end up being one such term. At any rate, the script learns what the user thinks is stupid, much in the same way that your e-mail client knows what Spam looks like. Then the program scans the text being displayed in the browser, and prevents the browser from displaying those phrases.
Here’s the StupidFilter FAQ, which states
…and…
So detecting irony or ignorance appear to be off the table.
I have to say seeing “ur duh”, or “whoosh” is enough to make me stop reading a post.
I guess having a filter like that would help if it were easy to turn off and on.
The trouble with filters is that even a few false positives are a real hazard.
Just today I found that my Outlook filters had dumped a letter from my secretary into the Delete folder. I had to add special rules to let people on my office list mail me things that otherwise look like spam, such as a message consisting of a single link to a web image.
I think it’s for real, but I’m unclear on how it’s meant to work - is it a plugin that stops you typing stupid things? (If so, who is stupid enough to need it, but smart enough to install it? And how long would they tolerate such an annoyance?)
-or is it meant to filter out stupidity from text you see online? - but how will it actually work? How will it distinguish the boundaries of the text that includes the stupidity? - identifying things like posts on a message board, or comments in a blog is quite trivial for a human - we can see who said what - where they started and where they finished, but it’s not as simple as that as a task for a computer - posts/comments in different places are all wrapped up and presented differently, so there’t no simple way to define what constitutes a coherent block of text contributed by one person on one occasion.
Heck, why don’t we just install it here and see how many posts it keeps out of GQ? 
Y r u agnst txtspk?
So he went like you know Molly fancies Pete and I went like no way!
The invasion of Iraq will be a great victory for the US.
Intelligent design is a science.
These two would be filtered.
These will continue to rely on the reader to apply critical thought before they can be rejected as potentially “stupid” things to say. Note the pathological case above.
As for Mangetout’s question about implementation, check out this section from the website:
That says to me that “we’ve got the filter, but we haven’t done the integration yet.” Website and blog admins will probably benefit the most, because they can pass all submitted comments through the StupidFilter, saving them valuable sanity points.