h a l f b a k e r yInexact change.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Markov Googler
Type in a search term and search engine makes up web pages | |
User would type in a search term and (optionally) a number for the Markov order. Google would then take the web pages that contain the search term and generate a page or more of Markov chains of the order specified. This could be applied to Usenet groups and news articles as well.
Depending on
the order and the number of pages it was using, the result would range from bizarre to almost intelligent -- and perhaps even funny.
(I've chosen Google to be the guinea pig since I like their search engine.)
The technology behind Google's great results
http://www.google.c...ogy/pigeonrank.html
You may have chosen guinea pigs, but google had another animal in mind. [bristolz, Apr 05 2002, last modified Oct 04 2004]
Markov Chain Gang
http://www.jwz.org/dadadodo/
Gives a description and examples of Markov chains applied to text. [mrouse, Apr 05 2002, last modified Oct 04 2004]
Markov Chain Defined
http://mathworld.wo...om/MarkovChain.html
Maybe you can make some sense out of it. [bristolz, Apr 05 2002, last modified Oct 04 2004]
The Daily Nonsense
http://www.lahosken...cisco.ca.us/markov/
Dissociated Press -- "News" based on Markov chains [mrouse, Apr 05 2002, last modified Oct 04 2004]
Proof of concept
http://lahosken.san...kov/goma_hectic.txt
I markoved together a text dump of the first 10 Google results for "hectic" and got this mess back. [lahosken, Apr 06 2002, last modified Oct 04 2004]
Dissociated Web
http://www.cs.washi...b/combine/foo2.html
As it happens, I was already planning to play with the idea of generating web pages through a Markovian process. This post gave me extra incentive. The synthesized page represents 5 top hits for "Markov Chain". It's a flimsy prototype, but kind of cute. [isohedral, May 17 2002, last modified Oct 04 2004]
Google Talk
http://douweosinga....projects/googletalk
Something along the same lines, which uses the first Google result for the previous n words. [codeman38, Dec 01 2004]
Markov thingy?
http://www.cgl.uwat...sweb/half/foo2.html
I don't know what this is, but it looks like it might fit here. [zen_tom, Oct 25 2006]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Annotation:
|
| |
Now there are 5 + votes and no negatives, and yet no annotations. mrouse, usually an idea this good prompts a cascade of annotations telling you how brilliant you are. I'm sorry you have not yet experienced this, and hope my feeble attempt restores some of the order missing in this little corner of the halfbakery universe. |
|
| |
I'd love to see this baked. |
|
| |
Maybe people are as surprised as I am at not seeing a couple of fish heads :-) I just like the idea for the entertainment value. |
|
| |
When doing a web search for the link on the left, I found out the author of the web page wanted to apply his dadadodo to AltaVista, so this might be "prior art" for applying markov chains to search engine results. |
|
| |
I'm having trouble picturing the output. If it's anything like the "Markov" ramblings I've come across on this site, I think it would lose entertainment value (presuming it had any) very quickly. Like, by the second paragraph. |
|
| |
Can anyone post a good example of this? Emphasis on "good". |
|
| |
The Daily Nonsense (link posted on the left) has its moments. It's supposed to change every day. One of the more famous Markov randomizers is called Bable, so if you combine it with Google you might end up with Boggle (or maybe Bagle, since this *is* the halfbakery after all) :-) |
|
| |
I must admit it is more humorous than I expected. Particularly the word-unit bit - reading today, I see this sentence: "I've fallen horribly behind in my life is not yet so empty that it's especially easy to do." It seems a fitting description for myself these days. |
|
| |
The character-unit stuff is much more difficult to sort through. As I read it, I thought about what it's like to read Chaucer in his original Middle English: "Pipen he koude and fisshe, and nettes beete, And turne coppes, and wel wrastle and sheete..." |
|
| |
Compare to the Markov: "Togetting somenting about he summed about the cynice loosed throught of condage, calle. Parked me her cand scread by much on Chuck?" |
|
| |
I believe you're describing a "blonde joke" generator. |
|
| |
Reminds me of a conversation simulator, MegaHAL (at megahal.net) that I've had a lot of fun with. |
|
| |
waugsqueke: I'm the guy who does that Chaucerian page. I agree with your "it would lose entertainment value... very quickly" statement. I never want to read more than a page of it per day. (So I never got motivated to speed up the program behind that page so that it could generate a page on demand :) |
|
| |
Hey, lahosken, I didn't realize you were the one to do that page, that's pretty cool. If you convert the program to work with Google and it impresses the judges, you could win $10K in a contest they're running. :) |
|
| |
After hammering a proof of concept together, I think I'm going to stop working on this. But I present some ideas for anyone else who wants to carry on. |
|
| |
Give more statistical weight to the frequencies of higher-rated pages? |
|
| |
Maybe the program should have a time limit. If Google returns 2,000 links, maybe you only want to slurp up 10 seconds' worth of results before markoving them. |
|
| |
I wish I'd known about lynx's -traverse option before I started. |
|
| |
I've seen a expirimental search engine that did something like this, and with a nifty flash interface even. I don't know the name, location, current status of company, or if it was in fact markov models. But it was a similar idea. |
|
| |
I'm going to bake this when I get some time. Shouldn't be to hard to do, the text that is. Throwing everything else in there will be hard. |
|
| |