Re: [agi] if yu cn rd tihs, u slhud tke a look
Interesting. I assume that OCR programmers already know about this. On 13/03/2008, Linas Vepstas [EMAIL PROTECTED] wrote: A bit of vision processing fun: http://www.friends.hosted.pl/redrim/Reading_Test.jpg --linas --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
Re: [agi] if yu cn rd tihs, u slhud tke a look
On Thu, Mar 13, 2008 at 8:35 AM, Linas Vepstas [EMAIL PROTECTED] wrote: A bit of vision processing fun: http://www.friends.hosted.pl/redrim/Reading_Test.jpg Interesting: is it possible to construct similar thing in audio form? It'll have to preserve some sequences of sounds, because we learn to parse sound sequentially, maybe separate pairs of sounds can be overlapped and shuffled if they are pronounced by different voices, because we can perceive different voices in parallel to a degree. It'll be a cool novel voice effect :-) -- Vladimir Nesov [EMAIL PROTECTED] --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
Re: [agi] if yu cn rd tihs, u slhud tke a look
On 13/03/2008, Vladimir Nesov [EMAIL PROTECTED] wrote: On Thu, Mar 13, 2008 at 8:35 AM, Linas Vepstas [EMAIL PROTECTED] wrote: A bit of vision processing fun: http://www.friends.hosted.pl/redrim/Reading_Test.jpg Interesting: is it possible to construct similar thing in audio form? Not to spoil the fun, but the human brain is adept at recognizing the same melody, whether its whistled, performed by an orchestra, or sometimes even just beat out with knuckles on a door. At the imperceptible level, there are catalogues of audio tricks known to sound alike to the naive ear, and these were employed by the designers of things like ogg, mp3, etc. The above is one of the few that I've seen that crosses the boundary of optical to linguistic processing. --linas --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
Re: [agi] if yu cn rd tihs, u slhud tke a look
On 13/03/2008, Bob Mottram [EMAIL PROTECTED] wrote: Interesting. I assume that OCR programmers already know about this. Traditional OCR tries to recognize one letter at a time, together with guidance from a spell checker. For this example, the spell checker would barf, so OCR might get all the letters right but would have none of the words as being dictionary words. --linas A bit of vision processing fun: http://www.friends.hosted.pl/redrim/Reading_Test.jpg --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
Re: [agi] if yu cn rd tihs, u slhud tke a look
One thing worth noticing is that it looks like this effect only works provided that words with three letters or fewer are not garbled. I think what this shows is that there is a statistical element to reading. So provided that the beginning and ending characters are correct, and what's in between contains some of the characters that you would expect to find in that word you can still read it. On 13/03/2008, Linas Vepstas [EMAIL PROTECTED] wrote: On 13/03/2008, Vladimir Nesov [EMAIL PROTECTED] wrote: On Thu, Mar 13, 2008 at 8:35 AM, Linas Vepstas [EMAIL PROTECTED] wrote: A bit of vision processing fun: http://www.friends.hosted.pl/redrim/Reading_Test.jpg Interesting: is it possible to construct similar thing in audio form? Not to spoil the fun, but the human brain is adept at recognizing the same melody, whether its whistled, performed by an orchestra, or sometimes even just beat out with knuckles on a door. At the imperceptible level, there are catalogues of audio tricks known to sound alike to the naive ear, and these were employed by the designers of things like ogg, mp3, etc. The above is one of the few that I've seen that crosses the boundary of optical to linguistic processing. --linas --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
Re: [agi] if yu cn rd tihs, u slhud tke a look
I reckon that the shuffled words (meaningless and low probability) trigger an internal representation that is close enough to the meaning_full_ representation to be correctly classified. One part of this triggered internal representation is about WHAT is present, the other part about WHERE these are present. The WHERE is a little different from the original word, but enough to trigger it. In a bayesian framework, this is extremely trivial, although the brain probably does it using some physically practical heuristic implementation. Durk On 3/13/08, Bob Mottram [EMAIL PROTECTED] wrote: One thing worth noticing is that it looks like this effect only works provided that words with three letters or fewer are not garbled. I think what this shows is that there is a statistical element to reading. So provided that the beginning and ending characters are correct, and what's in between contains some of the characters that you would expect to find in that word you can still read it. On 13/03/2008, Linas Vepstas [EMAIL PROTECTED] wrote: On 13/03/2008, Vladimir Nesov [EMAIL PROTECTED] wrote: On Thu, Mar 13, 2008 at 8:35 AM, Linas Vepstas [EMAIL PROTECTED] wrote: A bit of vision processing fun: http://www.friends.hosted.pl/redrim/Reading_Test.jpg Interesting: is it possible to construct similar thing in audio form? Not to spoil the fun, but the human brain is adept at recognizing the same melody, whether its whistled, performed by an orchestra, or sometimes even just beat out with knuckles on a door. At the imperceptible level, there are catalogues of audio tricks known to sound alike to the naive ear, and these were employed by the designers of things like ogg, mp3, etc. The above is one of the few that I've seen that crosses the boundary of optical to linguistic processing. --linas --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com