Google is turning to the public (or rather, web developers) for help testing its next-generation search infrastructure, code-named "Caffeine." The changes have to do with the way Google crawls the web and indexes content, so you shouldn't notice any changes int he search engine interface.

In order to try out the new version, visit www2.sandbox.google.com, and start searching. You can share feedback with Google by hitting the "Dissatisfied? Help us improve" box at the bottom of the page and sending a message with the word "caffeine" in it.

Overall, the search results look pretty similar, but there are a few minor changes. Some items may be ordered a little differently, while the text descriptions of some pages look different.

If you aren't particularly interested in testing the new technology to help Google out, here's another reason to give it a try: As far as I can tell, there are no ads on the Caffeine interface. I've conducted a few dozen searches, and so far I haven't seen a single sponsored result. Of course, that'll change by the time the new code is integrated into Google's main product. But it's nice while it lasts.

What other changes have you spotted? Let us know in the comments.

DocMazy is a novel kind of search engine exclusively dedicated to finding documents. Instead of returning websites, it digs up PDFs, DOCs, spreadsheets, PowerPoint presentations and text files. DocMazy was designed to find information that gets buried in other search engines because documents often don't score well in search engine algorithms, or aren't indexed by search engines at all.

Once you find the doc you're looking for, you can download it or view it online. DocMazy uses iPaper, which opens a preview without making you leave the results page. All in all, it's a pretty nice tool if you're looking for the kind of longer, more detailed information that often isn't published in its entirety on a webpage.

Google Squared, the new search tool from Google Labs, puts search results into a convenient spreadsheet format designed to help you get the basic facts about whatever you're searching for without clicking through to too many different websites. Search for "horror films," for example, and you'll get a list of movies with info like Author, Director, Cast, and Running Time. It's a great layout that's easy to read at a glance, but the results still need a lot of work.

I tried a search that I thought was right up Google Squared's alley: "current New York Times bestsellers," and the results were a jumble of categories, but no books. "Current box office" returned some recent movies, and some old ones that did really well when they came out. These are the kinds of searches Google Squared could hit a home run with, but it doesn't.

The example searches prove how well Google Squared could really work. Check out some terms like "roller coasters" and "U.S. Presidents" to see some nice glimmers of potential. The ability to customize your own square when no results are returned will also probably lead to much better results.

$timator is a site that purports to estimate the true monetary value of your website. Simply enter a URL, and $timator goes to work, analyzing things like search engine placings, back-links, and site structure to come up with an ultimate dollar value.

When I tried it for my 5 year-old personal blog, it came up with a relatively modest $1979, which to be honest feels a little generous (but I'll take it). However, Download Squad is apparently worth well over $31 million, which seems more than slightly absurd, particularly considering that based on this screenshot google.fr is valued at just over $18 million.

These "your website is worth x amount of money" have to be taken for what they're worth: a fun tool to play with and dream about how you could buy a few extra cases of beer if you were to sell your site. If Download Squad were for sale, can you imagine the party we could throw if we sold it based on $timator's valuation? You're all invited.

Visual search engine SearchMe is out of private beta and now has two nifty new features: Media search and stacks. Stacks is a fancy way of saying bookmarks, while Media Search lets you interact with online videos and images all from the SearchMe interface.

While traditional search engines like Google and Yahoo! present you with a list of text-based results, SearchMe shows screenshots of web pages matching your query. For example, searching for "Download Squad" would bring up a screenshort of our home page. When you click on the image you can visit the original page.

Now you can also click the video or images tabs and pull up screenshots of YouTube videos or Flickr images. When you click the play button you can even watch a video without leaving the SearchMe page. So you can flip through videos much like you would using the PicLens browser plugin, but without installing anyting.

The Stacks feature lets you organize collections of bookmarked pages. For example, you could create a list of your favorite tech news sites with thumbnails for easy reference. You can also share your stacks via email or social networking sites.

Text based search engines are all right if you're looking for text-based web pages. But Microsoft, Yahoo!, and Google all have features that let you search for images or videos, because let's face it, that's what you're looking for on the internet much of the time. But the major search engine's approach toward image and video search feels a lot like the old fashioned approach to text-based search. You get a list of results, a brief description, and a link to the original source. If you're lucky, you may get the ability to watch a video or view a larger version of the thumbnail without leaving the search page, but that's about it.

Viewzi is one of a handful of new web sites that's aimed at making visual search a little more... visual. Like visual search engine SearchMe, Viewzi lets you scroll between screenshots of web pages. But it doesn't stop there. Viewzi offers over a dozen different ways to view search results. You can pick your favorite view and then conduct a search. Or you can rely on Viewzi to automatically detect the type of information you're probably looking for and select a view for you.

For example, if you're searching for a song title, Viewzi will move the MP3 search view to the top of the list so you can find and listen to songs. If you search for "weather in Philadelphia," Viewzi will recommend a weather view or a news view, which shows associated content like a weather forecast or news stories related to your search.

Viewzi launched a public beta today, which means the service, which we first mentioned last summer is now open to anyone. But it is still a beta, and some search views work better or load more quickly than others.

There are plenty of image search engines on the web. Enter a text based search term, and Google, MSN, or Yahoo! will spit out a series of pictures it thinks match that query. But TinEye takes a slightly different approach. This image search engine, currently in private beta, lets you upload an image from your desktop and search the web for identical or similar images. You can also use an online image as your starting point.

The concept is pretty cool. If you're looking for an image, why should you have to enter text? But the truth of the matter is it takes a bit longer to search for something using TinEye than a text based search engine. And it's a lot harder to get accurate results. If you upload a popular image that's been used on a bunch of web sites, you should be fine. But if you upload your hand-drawn image of Angelina Jolie naked, don't expect to find any high resolution images matching your query. Not that we tried.

But seriously, we uploaded a few images that have been posted here at Download Squad, and TinEye didn't spit out a single match, which means it doesn't even indext this site. So while the service claims to be searching a whopping 487 million images, it doesn't exactly cover every site on the web.

Still, we'll be keeping an eye on TinEye. If the service can improve its index and do a better job of finding similar images, and not just identical ones, it could actually become useful.

[via WebWare]

Whether you use Twitter or not, you might want to check out Tweet Scan, a Twitter-based search engine. Basically, what Tweet Scan lets you do is search Twitter's public timeline for snippets of text. There are at least a few reasons this could come in handy even if you don't use Twitter:

1. You can find out what people are saying about a topic, web site, political candidate, etc right now
   
2. Twitter users may only be able to post 140 characters at a time, but they often post links to longer articles with more information, so you can treat Tweet Scan as a sort of blog/news search engine
3. Having trouble accessing a popular web page? Just type it into Tweet Scan and if a major web site or service is down, odds are other people are complaining too

If you are a Twitter user, Tweet Scan is also a quick and easy way to find other users with similar interests.

[via MakeUseOf]

Ever wish Google would show you previews of web pages before you visit them? RedZee may not load search results as quickly as Google, but it does let you preview a web site before you click through.

When you enter a search term, RedZee goes to work trying to find web pages matching your query. RedZee then arranges screengrabs of those web pages on a sort of wheel. Drag your mouse left or right to cycle through the pages. Double click on one and you'll be taken to that page. Sort of.

What actually happens is RedZee keeps a toolbar at the top of your window while loading the new page. That comes in handy, because you can easily open multiple pages and flip back and forth between tabs, all within a single web browser tab. But it also means you lose a tiny bit of screen real estate. And if you want to copy the page URL to your clipboard, you're pretty much out of luck.

[via Go2Web2.0]

Sometimes it's hard to find what you're looking for online. You go to Google, type in your search terms, and a few million results pop up. How are you ever going to find what you're looking for that way? Sure, if your search terms accurately described what you were looking for, the top results should be relevant. But sometimes it's hard to find the right search terms.

That's where SortFix comes in. You can use this site to search Google, Yahoo! or DMOZ. After you enter your search words, SortFix will show your results on the lower half of the screen. On the top you'll see a list of associated words. You can drag words that are relevant to the "Add to Search" box and the words that you want to ignore to the "Remove" box. Click Search again and you should get more accurate results, plus some new terms that you can add or remove.

Overall, you probably spend a little longer looking things up using SortFix than you would if you just went to Google. But in the end you might get more relevant results. Check out a cute little demo video showing how to search for homemade ice cream recipes after the jump.

[via Go2Web2]

We hear a lot of talk about the "Semantic Web" these days. The idea is that search engines like Google are inherently dumb. Sure, they do a pretty good job of returning accurate results to keyword searches. But because id doesn't actually understand natural language, you can't ask Google a straightforward yes or no question and expect an answer. Just search results.

Of course, for the most part, Google does a pretty good job of giving you what you're looking for. That's why Twine's Nova Spivack told us recently that his company decided to build a semantic social networking site rather than a search engine. But startup True Knowledge has its sights set on Google, Yahoo, and the other big names in search.

The True Knowledge search engine is currently in private beta, so you can't really check it out yet. And that makes us take everything in this demo video with a grain of salt. For all we know, True Knowledge can still only answer questions about J. Lo appropriately, but doesn't know the difference between Jennifer Lopez and Jennifer Beals.

That said, True Knowledge does look pretty interesting. We like the way that the web application does more than spit out a bunch of relevant web pages, but rather gives you an answer and then shows its work.

When we first told you about peer-to-peer search engine Faroo last week we promised to reserve judgment until we had some time to play with it. Now that we have, we have to say we're about as unimpressed as we expected to be.

The idea behind Faroo is interesting. The site indexes the web without maintaining a centralized database. Rather, users install a client on their PCs (Windows 2000 and up), and Faroo tracks the web sites those users visit.

When you go to perform a search, Faroo will only search sites that real people have visited. This is both the service's strength and its weakness. You don't get a lot of junk sites this way. But since Faroo is in private beta with a limited user pool, you don't get many sites at all that way.

For example, a Google search for the words "Download" and "Squad" pull up over 9 million results. Not all of them are relevant if you're looking for our web page. But Faroo pulls up just 5. More might not always be better, but at least it's not too few. If Faroo catches on and gets an enormous user base, perhaps this search engine could be great. But as with any social/user-generated software, there's a chicken and egg problem. It's hard to market a product that has a small user base, because it just doesn't perform very well yet. But it's hard to attract new users with a product that doesn't deliver all it promises.

On the bright side, the client doesn't seem to eat much RAM, so you could always install it now, go about your business and check out the search engine feature in a few weeks or months. We've got 10 invites to give away if you want to give it a try. Just be one of the first ten people to leave comment on this post. Remember, we've just got ten to give away, so if you're number 11, you're out of luck.

Update: We're all out of invites, but thanks for playing.

Ever wish Google would let you see what a website looks like before you click a search result? Yeah, neither did we. But Exalead has solved a problem that may or may not exist by putting thumbnail previews of websites next to search results.

Overall, the search engine performs reasonably well. There's a nice "related term" feature that lets you narrow your search results. And you can search the full web or just blogs. But the way we see it, there's two problems:

1. The thumbnails are far too small to really tell you anything. If you're searching for a product, you might have enough info to determine if you're looking at a web store or review site. But you might not. And for more esoteric searches, good luck.
   
2. The site includes 2-3 sponsored results at the top of the page. That would be fine for text results, but with images, your browser window can easily show 3 sponsored items and just one genuine, unbiased (we hope) search result. If you have to scroll down to get to the good stuff, people are just going to choose a different search engine.

In other words, the pictures aren't big enough to be useful, and they may also be too big to provide you all the information you want at a glance. The simplest solution would probably be to use fewer paid links, or to make the images for sponsored results a bit smaller.

The search engine does let you turn off the image preview, but if you do that, there's really not much reason to use Exalead over competing products.

Exalead will demonstrate its search engine at the DEMO conference next week.

[via Silicon Alley Insider]