These are some areas in which nobody has innovated enough, thus leaving it open for anyone to basically do a google maps/gmail:

1. Image Search – Turn web image search into something comparable to the quality of flickr. Add the ability to match images. Integrate creative commons. Improve the interface.

2. Directory – Add ajax to make it more worthwhile, add suggested catagories, and automate it. Turn it from a "finding a web page on a specific topic" to a "surfing the web"

What do y'all think? What existing tools do all the major search engines have but that need major improvements in your opinion?

1. If they integrate Creative Commons and the hosts images are Copyrighted (All Rights Reserved). Wouldn't this cause issues.

==="which nobody has innovated enough"

I differ on that statement

others are innovating too..take for e.g Riya , which has facial recognization within its image search.. their algo's are way ahead of the curve from most image searching folks..

what google needs to do is tie down all their existing services properly.. writely, CL2, SS, Gbase, Picasa, Maps, gmail and create a std API which permits generic users to superimpose data from on app onto the other.. this will create a platform dependant space for most Internet users..

for eg, If I update my calendar with information (or accept a schedule), I would like the venue to be directly interfaced on my map ...

1. I asked Matt Cutts about that. C.f. http://blogoscoped.com/forum/59472.html#id59616 Also in previous threads (e.g. on PicasaWeb) I suggested some features like this. So I totally agree. The only thing is that this imposes a lot of AI & is maybe to big to do with present technology.

2. Do you mean something like the Livesearch of Alltheweb.com?
http://livesearch.alltheweb.com/?ek=1#&p=google%20blogoscoped

btw "Add AJAX to make it..." => "Add AJAX *and* make it..."

Sohil: I mean have an option to show only creative commons images or an indication in the result set that an image is creative commons

/pd: when I said "anyone", i really meant big players. It would count if google, yahoo, or ms *bought* them and used them in thier real image search/directory

Art-One: (#1) yeah I read that and that's kinda what prompted this post

(#2) I'm not saying this stuff has to be updated in a specific way, i'm just saying that these areas of search seem to be neglected in general and could do with some innovation.

Caleb: (#2) I fully agree...

Caleb, Gotcha. But Google already has a Creative Commons Search

http://www.google.com/search?q=India&as_rights=(cc_publicdomain|cc_attribute|cc_sharealike|cc_noncommercial|cc_nonderived)

not for images, and it's very hidden.

Flickr is more appropriate for finding CC images as people declare their licensing. Google is not able to find if an image is CC-licensed or not.

they could do it based on the page (like for thier CC page search). Does cc have a provision for making images in a page creative commons?

I'd love to see an image comparison engine with web wide scope. Sure, we've seen prototypes working on a smaller set of images, but can anyone pull it off in huge scope? Is this simply too much processing power even for Google, to compare images?

There is also no *terrific* web-wide video search engine. A search engine with great scope and usability. I would compare current video search engines to about the quality level online maps were before Google Maps. Then Google Maps came along, and now most major online maps look like Google Maps. Why hasn't Google gotten into web wide video search yet?

Sidenote: Hmm, maybe one could implement a meta CC image search running on top of Google.
1) Search the Web using Google's CC search options
2) Grab the domains
3) Search Google Images using the site:domain.com operator on CC domains only
4) Present CC images to the user...

Flickr's CC search is nice, but it has some problems, IMO:
- Flickr is often very slow
- Flickr doesn't let you search through all CC licenses at once, you need to decide exactly which license you want
- It's not web-wide...

I agree Philipp. There seems to be a big conflict between user submitted content and web content. Google video & flickr are exclusively submitted content. Wny not add web stuff in there too? It's ridiculous to expect web page owners to submit thier content to search engines; why should images/video be any different?

<< It's ridiculous to expect web page owners to submit thier content to search engines; why should images/video be any different? >>

Many people I know use services like YouTube and Google Video simply because it saves them the hassle of hosting the content in the first place. Similarly with Flickr.

How many people with websites actually host videos anyway? It's easy to setup a blog/website these days but to host movies in a file format that's viewable by 99% of your visitors (and embed them in your website) is much harder. Google Video and YouTube make this much easier to do.

(Of course, images are different – and Google Images already crawls existing websites, unlike Flickr.)

Anyway, back to the topic: Unless we can add some kind of CC header to images and videos themselves, it would be hard to determine whether they're actually CC controlled. For example, you can't rely on an additional file (i.e. like some kind of robots.txt file for Creative Commons) because if someone downloaded the image/video, they'd have to download and host that file too. It just wouldn't work.

Finally, one flaw with Philipp's suggestion of implementing "a meta CC image search running on top of Google" is that Google Images often returns hotlinked images from other websites. Also, just because an image is linked to from a CC website doesn't mean it's actually CC licensed.

If you want to search video's on the web there is a search engine that does the job: http://www.alltheweb.com/search?cat=vid&cs=iso88591&q=google&rys=0&itag=crv
(It's the experimental search engine of Yahoo.)

I find mostly what I need with this but I'm not expert enough to find out of the overall quality is good or not...

> Finally, one flaw with Philipp's suggestion of
> implementing "a meta CC image search running
> on top of Google" is that Google Images often
> returns hotlinked images from other websites.

Well, you could exclude such hotlinked images. And I'm not saying I know how to program such a thing with any realistic return speed... I mean just querying for say the top 100 Google web domains takes 10 API requests, and then you gotta do screenscraping on 100 additional Google Image search results... and then you gotta do further analysis of the HTML contained on those pages! Would probably take endless, unless you have straight access to the data or Google finally gives us a kind of bulk-API.