![]()
Visiting guest. Why not sign in?
|
|
Reply to Message
Not registered yet?
The AI Depot has a focused community of friendly users. Rather than let anyone abuse the site at the brink of promiscuity, we prefer to let only those with an active interest participate... this simply requires registering.
Why not sign up!
Joining the site's community is completely free. You can then post messages freely, and customise your personal profile at will. Specific privileges will also be granted to you, like being able to access printer-friendly articles without restrictions. So, why not register?
Parent Message
'Dumb' Spell Checkers
Some of you may have had an email forward containing a poem, Ode to a Spell Checker, which goes something like this: "Mistake eye cannot sea, I've run this poem threw it, I'm sure your pleased too no." From then on, the Herald Tribune goes on to assume that humans are needed, even with tasks as trivial and repetitive as spell checking (Why Humans Are Necessary, Even With Spelling Checkers). Quite frankly, instead of matching single words, a Natural Language Processing solution would fare much better. That said, it wouldn't be perfect... Experts are currently still torn between time consuming expert systems -- based on huge sets of hand-crafted rules -- and learning approaches based on statistical analysis -- which are still in their infancy. |
|
A little
What this article really highlights is that giving a program a small amount of narrowly defined "intelligence", in this case the ability to spell thousands of words correctly, can lead to users assuming too much of the software because they expect something that can spell must have some level of general intelligence or "common sense". The lack of "common sense" in standard expert systems was the inspiration for Doug Lenat's Cyc Project, which has gone on to show what a huge task it is to capture the everyday knowledge that we carry around with us and use on a daily basis. Unfortunately, I think we will probably see more accusations in a similar vein as limited AIs are deployed and misused by the public. Not that this is a very bad thing, it's just going to take some time before we start making more "natural" AIs that fail in ways that people can forgive more easily. Personally, I've always been drawn to the machine learning approaches, e.g. neural networks, because their performance degrades nicely when faced with unfamiliar situations. -RoB- PS: AI-Depot's spellchecker reports 8 false-positives in this text, including "spellchecker" (twice). :-) |
|
Spaling Misstooks
The Cyc system will take years! I'm convinced that a much simpler data-mining tool that learns grammars from correct text would be fairly simple to produce, but I haven't yet got round to looking into it. As for the site's checker, it's just a static checker based on the standard i-spell. No word dependencies are taken into account, so yes, it does give the odd error. However, it's quite good at picking out my most common speed-typoes (inverting two letters and missing one out). |
|
Assumptions
> users assuming too much of the software True! That's probably the biggest cause of the 'AI effect'... AI comes across as an 'over-rated' technology that can't perform; it can, just not to your expectations, yet. I think that applies to computers too. Sometimes I just want to yell: "Just pick up a pen and a piece of paper you moron!" (slightly edited ;) But I guess it's this sort of things that drives the business, by producing stuff that people expected in the first place -- and charging them extra for it! |
|
Back to the Artificial Intelligence Depot.