January 31, 2008

Search Comes Full Circle?

Reading the search headlines these days, it may seem like we’ve gone a long way to get right back where we started: human-generated search. Well, not exactly where we started; these new breed of search engines aren’t human-powered in the same way DMOZ and the original Yahoo were, they’re algorithmic search engines that have been human-enhanced by allowing searchers to rank or vote on results, and even to tag or comment on them—much like social bookmarking sites like Digg and Reddit do.

And they are growing in popularity, with some of the biggest names in search behind them. Jason Calcanis, entrepreneur poster-boy and SEO public enemy #1, recently introduced an update to Mahalo , his human-powered engine, that adds aggregation of user profiles and pages from various social networks . Matt Cutts hinted that Google was integrating social interaction into results and we’re beginning to see Google test it. And Google’s best friend, Jimmy Wales, is making headway with Wikia Search, his admittedly “poor” but improving search engine that integrates the philosophies of Wikimedia and user-generated content.

So are human-enhanced search engines really the future? And if they are, is that a good thing?

As with any search engine, the first criterion that needs to be addressed is relevance. Do social search engines like Sproose, which promises “user-improved” results, really provide better results than regular algorithmic search engines? Looking at most social search engines, including Sproose, Mahalo, Wikia, ChaCha, EarthFrisk and Eurekster’s Swickis, it seems that the answer is generally “no.” Social search engines, or any search engine counting on human participation, need a critical mass of said human participation, which none of these engines seem to have (at least, as of yet). For the most part, each engine returns nearly identical results to Google, with the exception of some highly popular search queries.

The second thing to look at is the trustworthiness of the results, or—to be more frank—the excess or lack of spam therein. Do social search engines prevent SEO spam by giving more power to their audience and less to SEO professionals, or do they open themselves up to more spam by allowing regular (registered) users to directly influence rankings? In an ideal world, social search engines would police themselves from spam the same way social bookmarking sites and networks do—by counting on their user base. Social search would then solve the problem of aggressive SEOs pushing irrelevant sites to the top of the SERPs and create an ideal, democratic system of ranking search results.

But it’s not an ideal world. These social engines have such small user bases that nearly all search spam that exists in traditional engines gets through to them as well. And on pages where users have contributed, such as gambling and SEO queries, social spam unique to these new breed of engines is rife. If anything, these engines are the worst of both worlds from a searcher’s perspective.

But that doesn’t mean the idea of social, human-empowered/enhanced search should be abandoned. The theory of allowing users to remove obvious spam and to optimize results pages based on true semantic knowledge (they know what they’re looking for, after all) is the light at the end of the tunnel that can truly propel search to its idyllic usefulness. But it is unlikely that a minor player will get there. Social products survive on a mass of willing users. Google, Yahoo and Live Search have those users. No one else does, and it is unlikely that others will acquire them. For social search to really work, Google is going to need to push its experiment beyond Google Labs (which it seems they are doing)—or Yahoo and Microsoft will need to step up to the plate. One thing is for sure: whoever gets there first will be tomorrow’s search leader.

Posted by Eli Feldblum at 5:53 PM | Permalink

May 17, 2007

Google Experimental to Offer Sneak Previews

At the Searchology event yesterday, Marissa Mayer, Google's VP of search products and user experience, introduced Google Experimental, a new tool that will allow users to opt in to various tests that Google is doing.

It has been Google's policy to roll out potential new features or interface tests to small groups of users before going live with all users. This of course leads to frustration for SEOs and other Google-watchers who want to see what these new features and interfaces look like.

Now, those leading-edge users can sign up for these tests at the Google Experimental site, and they will be added to the test group. Google will continue to conduct random tests with users, since the self-selected group using the new tool would skew results, but Mayer said she hopes that that group will be able to provide Google with valuable feedback as well.

The first tests available on the site are timeline and map views, keyboard shortcuts, and left- or right-hand navigation options.

The concept of Views was introduced by Mayer yesterday as a way to extract structure from unstructured information. For the timeline view, for example, a user could see the history of the civil rights movement throughout history. With a map view, a user could plot out the locations of golf courses on the PGA tours. Other kinds of views will continue to be added.

Keyboard shortcuts allow a user to navigate search results with a keyboard, in much the same way that users have been able to navigate around Gmail. The navigation location preferences allow a user to test out a layout that moves the new overhead navigation to the left or right side of search results.

Google Experimental will not replace Google Labs, which will continue to house standalone applications and products to test. With Experimental, users need only go to the site to sign up, and then all searches on Google thereafter will include the test functionality or interface, until a user opts out of a test.

Posted by Kevin Newcomb at 1:41 AM | Permalink

October 2, 2006

Google's New SearchMash Test Site

Google's gained a new unbranded site called SearchMash where it plans to test user interface ideas without Google's brand somehow skewing the tests. Below, more about the site and comments from Google about it.

Currently, SearchMash allows you to perform a search and get web and image results presented side-by-side. It's similar to how A9 has long allowed side-by-side results, ironically a feature that A9 has made much harder to implement after a recent redesign over there.

Web results are presented in the main left-hand column after a search and seem ranked the same as at Google. Unlike Google, presentation is different. Results are numbered. Clicking on the URL line makes a box pop-up with options to:

  • open the listing in the current window
  • a new window
  • to see more pages from that web site
  • to find similar pages.

After the first ten results, there's a "more web pages" link at the bottom. Click on this, and you get another 10 results magically appearing on the same page, inserted below the first 10. You can keep going, adding 10 more results at a time.

It's pretty slick. Microsoft's Windows Live had a somewhat similar "infinite scroll" feature that allowed you to keep getting more and more results, as you went down the page. Unveiled in March, it was dropped in September for web results (it still works for image results) when Windows Live came out of beta, as Microsoft felt it slowed performance.

While A9 dropped so many features, "continuous scroll" is something it gained. Do a search there, and as you scroll down, more results keep magically appearing, 10 at a time.

Unique to SearchMash is the ability to drag-and-drop web search results. Click on the number next to any listing, and you can move that listing higher or lower in the search results. The number doesn't change after you move it. The feature also doesn't seem that useful. Far better would be a scratch pad-style feature such as Windows Live offers for image search. Being able to drag-and-drop web results into some type of collection area would be handy -- and it's something that Microsoft is promising.

Those are the features at the moment, which you can also find described on the site's features page. What you won't find is much about Google on the site. The About page doesn't mention them. You've got to go into the privacy page where you discover:

SearchMash is a website operated by Google Inc. The Google Privacy Policy describes how we treat personal information when you use our products and services, including information provided when you use SearchMash. In addition, the following describes our privacy practices that are specific to SearchMash.

So what's up with SearchMash. I fired some questions off to Google, and here's what I got:

Q. When did this go up from Google?

Very recently.

Q. Why are you doing it?

  • SearchMash is an experimental search site operated by Google. The goal of SearchMash is to test innovative user interfaces in order to continually improve the overall search experience for our users.  
  • The site does not include Google branding to help us gather more objective data about user response to new interfaces.  
  • There is no guarantee that the features tested on SearchMash will be seen on Google search. As with all of our experiments, one of the main factors we will consider is user response to the feature and how well it addresses their needs.  
  • This site is only a test and has traffic limitations so may be unavailable at times.

Q. Why is it not on Google Labs?

Google Labs continues to be a great site for Google to launch new products that may not be ready for prime time yet, frequently and quickly. In this case, one of the important factors we wanted to address was the influence that may come from Google branding. Creating a separate site will help us gather more objective data about user response to new interfaces.

OK, next some follow-ups and speculation. First, how can a site that no one knows about be useful to Google? Pretty much no one heard of it until the past day. As best I can tell:

So the site's going to have plenty of visitors, but all the wrong type, people who are the influencers or tech-heads or early adopters that Google's not trying to test against.

Remember, Google's been doing a lot of testing over the past year or so. Barry Schwartz just noted yet another sidebar navigation experiment yesterday. The experiments became so frequent and much discussed that I was begging Google in March to provide more official notice about what they were doing. Google's response to me was that announcing the experiments would skew the results.

Still, with everyone watching them so closely, experiments were quickly noted by the blogging community. That may have helped Google decide in April to blog itself about how it tries to test things against small groups. It even illustrated some of its experiments.

Now SearchMash gives Google an experimental playground, one similar to how AllTheWeb is supposed to operate for Yahoo, though aside from LiveSearch being launched there in May, Yahoo's not done much with AllTheWeb.

Google can play with weird stuff at SearchMash without worrying about "normal" users having the Google brand set up expectations. But how do those normals (or "mundanes" for you Babylon 5 fans) get to the site? From Google:

We have various methods for driving traffic to search and UI experiments that we run but we don't share details regarding the methodology to help keep the results as objective as possible.

A couple of guesses here. Google is likely (or will be likely) to divert people to the site in various ways, such as perhaps if someone uses an AdSense For Search box on a content site. It might simply push some people trying to reach Google to SearchMash (perhaps with some interstitial page warning them beforehand). It also gives them a site to put before controlled focus groups, where they might not know Google is behind it.

What about the skewing that will happen now that early adopters and the Google-obsessed will be all over SearchMash? They can be filtered out. If Google is directing certain groups to the site in various ways, it can then filter studies of user behavior to just those groups.

OK, one last thing. What about the idea that SearchMash will be the new place for Google to allow people to create custom search engines of their own, similar to Yahoo Search Builder launched last month or the older Rollyo or Eurekster Swicki services?

Garett Rogers last week wondered if the IndexBench trademark Google applied for recently was a sign that Google was planning custom vertical search engines. Now he wonders if SearchMash will instead be the place for this, after Google Operating System highlighted a Time Magazine article confirm that Google plans this:

Marissa Mayer, who manages search products, says the company has assigned more engineers to search than ever before and plans to release a new search tool that will enable users to design and build their own flavor of Google search, scanning just the sites they're interested in.

So yes, custom vertical searches are coming, likely more substantial and customizable than the long-standing Site-Flavored Google Search that's been out since 2004 and recently upgraded this year. But Google wouldn't say if it will be on SearchMash or not. So wait, watch and see.

Finally, the feel of SearchMash to A9 in many ways is uncanny. As I noted in my Amazon's A9 Becomes, Well, Sort Of Nothing post earlier today, A9 was an experimental playground for Amazon that seems to have lost its way after it lost its CEO Udi Manber to Google. I haven't heard back from Google on whether Manber is now running SearchMash. But seeing the side-by-side results that were a hallmark of A9, plus the infinite scroll similar to what A9 just rolled out (and what may have been in the works before Manber left A9), it sure feels like he's running a new playground search engine -- this time for Google.

Posted by Danny Sullivan at 7:26 PM | Permalink

September 27, 2006

What's Cooking in Search Engine Labs

Many of the major search engines showcase the projects they are working on in their respective research labs. Want a peek behind the scenes? Read on in today's SearchDay article, Behind the Scenes in the Search Engine Labs.

Posted by Chris Sherman at 2:25 AM | Permalink

July 20, 2006

Google Labs Launches Accessible Web Search

Google has launched Google Accessible Search, a stripped down version of the Google search results page that also serves up results to sites deemed more accessible to those who are visually impaired.

The design was created to make it easier and more effective for the blind and visually impaired. The results are a bit different, tailored to the visually impaired, to show more accessible pages in the results. Google told me how this actually works:

Google Accessible Search looks at a number of signals by examining the HTML markup found on a web page. It tends to favor pages that degrade gracefullypages with few visual distractions and pages that are likely to render well with images turned off. Google Accessible Search is built on Google Co-op's technology, which improves search results based on specialized interests.

It is also important to note that Google Accessible Search is currently ad free, conduct a search for ipod and you'll see no ads.

Matt Bailey has his take on this new Google product at his Accessibility Blog. Postscript From Danny: While I love the new offering, it's still disappointing that Google didn't provide an easy way to increase font sizes in the way that Big.com does.

Posted by Barry Schwartz at 1:09 PM | Permalink

February 22, 2006

Google Launched Google Research Blog

Philipp Lenssen located a new Google Blog, at googleresearch.blogspot.com, which will be a blog written by some of the smartest people at Google. In the opening blog entry Peter Norvig, Director, Google Research describes who works in Google Research.

So who are we? We're experts in machine translation who came here to work with the largest corpus of bilingual and monolingual text ever assembled. We're experts in machine learning algorithms who came to work on one of the world's largest computing clusters. We're researchers in natural language, vision, security, human-computer interaction, and a dozen other fields who came to help a user base of hundreds of millions of people. And we're working side by side with the engineering team -- not in a separate building or site. Some of us are launching projects on google.com this week and wearing pagers, and some of us are working on goals for the year 2020.

Yea, Google has about a hundred Google Blogs. There is the main Google Blog, the AdWords API Blog, Blogger Buzz, Google Base Blog, Google Blog (China), Google Blog (Korea), Google Enterprise Blog, Google Code, Google Maps API Blog, Google Reader Blog, Google Talkabout, Google Video Blog, Inside AdSense, Inside AdWords, Inside Google Desktop, and Inside Google Sitemaps. If you want one place to track them all, just use Phillip's Google's Blogs 'n More page. Oh, Phillip's "Google's Blogs 'n More " contains many of the Google employee blogs, I believe.

Posted by Barry Schwartz at 10:51 AM | Permalink

November 16, 2005

Matt Cutts Has a Cup of Coffee at PubCon Conference

Like he does at many of the Search Engine Strategies conferences, RustyBrick (Barry Schwartz) is blogging from the WebmasterWorld PubCon currently underway in Las Vegas. This post: Coffee Talk with Senior Google Engineer: Matt Cutts, offers a great Q&A style review (not an official transcript) of today's hour long session. Kudos to Barry for making it available.

Here are two of the more interesting Q&A's from the audience:

Q: CSS positioning? How does it affect ranking. A: Good question, I don't know. If your doing an include, it probably wont matter either way. In his mind, positioning text at top or bottom, is over rated. But try it.

Q: Google Analytics, can you confirm that Google will be using that data in the search engine? A: He cant confirm, but he can deny it. :) Matt as a Web spam team member, does not have access to this data. He wont even ask for it. If it becomes a concern, he will post it on his blog. People will always be concerned, so don't use it.

Postscript: Aaron Wall was also at the Matt Cutts session and he shares his overview here.

Posted by Gary Price at 4:52 PM | Permalink

July 28, 2005

Yahoo Hiring From IBM

We don't need no stinkin' Microsoft execs like Google, says Yahoo. We'll rob IBM! Yahoo Is Wooing I.B.M. Technical Talent from the New York Times looks at how Yahoo just picked up Prabhakar Raghavan, formerly of the much cited Clever project (and part of the foundation for Teoma), as head of research. He's not directly from IBM but comes to Yahoo via Verity. But another direct IBM hire has happened and Yahoo says more are in the works. This follows on Yahoo recently opening a new research lab at UC Berkeley, though they did lose their former Yahoo Labs head to Microsoft in April. Google, meanwhile, prefers to raid Bell Labs. See From Bell Labs To Google Labs for that.

Posted by Danny Sullivan at 9:30 AM | Permalink

June 30, 2005

Behind the Scenes at the Major Engines

All of the major search engines, to one degree or another, provide insights into what they're working on in their research and development labs. The quality and quantity of what's shared varies widely, but you can get a good sense of what to expect in the future by spending some time with what's available. Today's SearchDay article, What's Cooking in Search Engine Labs shows you where to find the best sources of inside information, both official and unofficial.

Posted by Chris Sherman at 10:28 AM | Permalink

June 1, 2005

New Compilation of Research Papers by Google Engineers

A new compilation with direct links to eight research papers (the actual papers are not new) by Google engineers is now online. All of the papers deal directly with Google technology. A link to the page can be found at the bottom of the Google Labs home page. The original "Papers written by Googlers" compilation remains online and is available here.

Posted by Gary Price at 11:55 AM | Permalink

March 31, 2005

New From Google Labs: Google Ride Finder

New from Google Labs is Google Ride Finder (GRF). This service uses GPS data to pinpoint and map (using Google Maps) the location of taxis, limos, and shuttle vehicles available for hire in 10 U.S. metro areas.

The 10 areas are: + Baltimore, MD + Chicago, IL + Cleveland, OH + Dallas, TX + Milwaukee, WI + New York, NY + Phoenix, AZ + San Jose, CA + St. Louis, MO + Washington, DC

For example, in the Baltimore/Washington Metro area Google is working with Super Shuttle (a company that takes people to and from the area airports) and just one taxi company in Alexandria, Virginia.

In San Jose, Google maps the whereabouts of taxis from Silicon Valley and Yellow Cab - San Jose and Checker Cab of Silicon Valley.

Remember, just because a taxi or shuttle van is only block or two away from your location doesn't guarantee they're going to come and pick you up but heck, this might help get the taxi/limo there sooner. More importantly, I think this is an interesting beta of location aware services that will soon become part of our daily lives. Google also provides a page where fleet owners can ask for more info about becoming part of the service.

Google also mentions that they're "working closely" with a variety of companies including Mobile Knowledge, Tranware, and Vettro. I wonder if they're also using some of the technology they got with the acquisition of ZipDash that Mike B. blogged about yesterday.

This service would also be useful to mobile web users. I'm guessing that the Google Mobile team has this on their "to do" list.

Google Ride Finder reminds me of another service called NextBus that does the same type of thing with bus and train location info for several U.S. metro locations. They even offer a web alert service.

It also reminds me of several air traffic info services that have been around for years. I'm not talking about the many sites that offer estimated arrival times of when a plane is due to arrive at an airport but the tools that also provide near real-time maps and other data (speed, direction, altitude) of all planes currently in an FAA database. You can see what I mean here and/or here (look for the quick track menu).

Posted by Gary Price at 11:55 PM | Permalink | TrackBack

March 10, 2005

Google Makes Suggestions in Japanese

Here's one for those of you who search in Japanese, or know people who do, a Japanese version of Google Suggest is now online.

Posted by Gary Price at 1:10 AM | Permalink | TrackBack

September 29, 2004

Questions from the Google Labs Aptitude Test

Both G.L. and J.B. point us to this blog post by Adam Rifkin that includes questions from the Google Labs Aptitude Test.

UPDATE: Google Has Posted the Test on the Google Blog.

Posted by Gary Price at 5:12 PM | Permalink | Comments (0) | TrackBack