Vector One starts a good discussion on why current search is actually pretty bad. His proposed Doodle is just the beginnings of an idea. The premise is: search is dumb, will staring at a "gajillion" (or another aptly named large number) really help me find what I'm looking for? How many of you go beyond the second, or maybe third page of results? And what do you do to get better results, try adding some more terms which may or may not actually help you get to your answer.
There are several possibilies that could easily be added to increase the power of search:
My trusted sources
del.icio.us and ma.gnolia.com both provide very easy ways for me to bookmark and share interesting sites. Bookmarking a site is putting a stamp "I like/trust this source of information". Therefore, when I search, I should be able to limit my results to this set of trusted information. I've already selected these as sites that I want to go back to for information, but currently I'm forced to go to each of these sites individually and look for the answer to my questions.
Google supports an
insite: option. It would be easy to say, search for a term within any del.icio.us tag or bundle. del.icio.us itself could build this right into its own site. For example, say someone wants to find out about "Dog Fleas". Putting this into Google may or may not yield results you actually care about. But you have bookmarks for a bunch of forums and sites you like for canine information. Google should use these as your most trusted sources.
Categorization of results
Google has started addressing this problem by having "split results". For example, a search for magnolia shows results pertaining to the bookmarking application first, then a horizontal line, and then magnolia the plant (why the web application is first is a whole other post). This is a good start, though I don't think a horizontal line really conveys the split in regimes.
A better solution would be to provide the user with some kind of categorization, perhaps on the right-hand side, based on the broad possibilities. When I search for "magnolia", it would display: "bookmarking application, tree, japanese artist...". I could then select any of these to narrow my search within that regime.
Another perfect example would be when you search for a person. My own example "Andrew Turner", turns up the broader categories of:
- Andrew Turner: Isle of Wight MP
- Andrew Turner: Engineer and Programmer interested in Geolocation
- Andrew Turner: Professor of Biology
- Andrew Turner: Game and general programmer in Massachussettes
- Andrew Turner: Carburettor Specialist
- and so on...
Anyways, the point is, these generalizations could be presented to help narrow down that search. Currently, you're relegated to having to try and enter more terms about the "Andrew Turner" you're looking for, hoping this extra term is on the same page/group as the rest of the information.
Additionally, within a person or site search, using XFN relationship tags like
rel="me" the result could pull up my blog, Flickr, LinkedIn profile, various projects, or even my other relationships like my fiancee's blog, my friends' blogs, or people I've met.
The power is in your hands
What is great about these suggestions, and the rise of mashups and services is that anyone can create this solution, with very little cost. Make a delisearch, or a relasearch. Just send me the links so I can start using it this afternoon.