Better Search Engines
prostoalex writes "Scientific American is seeking better Web searches. They report on all sorts of innovations happening outside the Google-Yahoo-MSN zone that the press is usually reporting on, including GPS-enhanced searches from University of Maryland, Shape Retrieval and Analysis from Princeton, musical search engine from New Zealand Digital Library Project, and some of the projects that A9 and Ask.com have been working on."
a user can record a query by playing notes on the system's virtual keyboard. Or he or she can hum the song into a computer microphone.
I tried that, but I was so out-of-tune the search engine returned all songs from Britney Spears.
Rock that crushes, Paper & Scissors that don't matter.
Is just some better work done on recognizing essentially similar documents. Like, if I perform a search, and 40% of the returns are the same wikipedia article copied to different sites, it would be nice if the search engine could only show me one (wikipedia). Or, like, if I'm searching for some kind of error I got while using Linux. Most of the returns I get will be various old Linux mailing lists, but only some of them will be relevant to my problem. There must be some way the search engine could logically organize them for me so that I could more clearly identify that block of returns that is most applicable to my problem of the moment.
It's because all pop music sounds exactly the same these days. From the perspective of a search engine, it *IS* all just songs from Britney Spears.
You can do this in google: searchterm1 searchterm2 ~bogus The tilde will look for synonyms. You can see which ones hit back by reading the bold results which are neither searchterm1 or searchterm2. I use ~howto and ~cheats often.
Rule of the open mind
People who are resistant to change cannot resist change for the worst.
The number one search engine feature that would make my life infinitely easier would be precise proximity operators in search engine syntax.
(For those who don't have a clue what I'm talking about, LEXIS-NEXIS, among others, allows you to run searches like foo w/5 bar (the word "foo" within 5 words of the word "bar"), or even foo pre/5 bar (the word "foo", followed, within five words, by the word "bar". Good proximity engines allow you to search not only within x words, but also to order terms, to specify root words within terms, etc.)
It would be great to have people reviewing and whitelisting page results, but that takes human interaction. Implementing precise proximity operators, though, can give you nearly the same benefits without any of the human cost.
Many people here have suggested eliminating ad text from search results, but if history is any indication, any algorithmic system that we can come up with to do so will be circumvented pretty quickly. The one way to fix this is to allow me to say that I want the word "modperl" within 10 words of "solaris", rather that just specify any page that contains both terms. That will get rid of 95+% of ads right away.
Surely, with all the bright people at Google, this is something that they can figure out pretty easily.