New Sequoia-backed visual search engine SearchMe launched a bunch of new features today - new video and image search engines as well as a new visual bookmarking tool called stacks.
The main new feature, stacks, allows users to bookmark and group sites and share them, visually, with others. To create a stack, you simply drag results into a newly created stack. See the how to video below, and here is a sample stack of companies that launched at the TechCrunch40 Conference last year. You can see more public stacks here.
I’ll be the first to admit that the first (private) release of SearchMe was a little rough around the edges. The results look great, and it’s fun to scroll through them like albums in iTunes, but the relevance and ranking wasn’t so hot.
Relevance and ranking is getting better, though. It’s the focus on the company now, says CEO Randy Adams. And the effort is being led by new VP of Research Mike Mathieson, who joined the company three weeks ago from Yahoo, where he was the director of engineering for web relevance.
SearchMe is one of only a handful of companies that indexes the entire web, so they’re serious about evolving into a big search player over the years. Search volume is up to 100,000 - 200,000 queries per day, says Adams. so they must be doing something right. Some users just want the quick text search results that they’re used to, and SearchMe’s visual results just slow down the process. But others (like my parents) like seeing the page before clicking on it.
New Sequoia-backed visual search engine SearchMe is just starting to send invitations to their private beta, which launched last week. The company says there are 30,000 people now on the waiting list. But if you want to get in now, just click here and enter your email. The first 1,000 people get in immediately.
Mountain View based SearchMe has been around since 2005 and has raised $31 million from Sequoia, DAG Ventures and Lehman Brothers. But until last weekend when I met founders Randy Adams and John Holland I knew next to nothing about them. It now joins Mahalo as one of Sequoia’s big bets in search.
I say “next to nothing” because I actually did write about them before. In January 2007 the company launched a test product called WikiSeek that returned results only from Wikipedia and sites linked from Wikipedia. At the time Adams said WikiSeek was just a test product for the technology they developed. Now, over a year later, their ready to put up their main site.
SearchMe goes into private beta today - sign up on the home page for an account. The main difference between SearchMe and other search engines is that it returns results primarily in a visual format, via an image of the result site. The results are displayed in a way that is similar to browsing through albums in iTunes - see the following videos to get an idea of how it looks:
Today SearchMe has indexed just a tiny portion of the web - about 1 billion pages. But they are quickly adding to the index, and say that what they’ve grabbed so far is suitable to show off their technology. In the live demo I saw some results that were great. Other searches returned only so-so results.
Categorization and Vertical Search
SearchMe isn’t all flashy graphics. They are also auto-categorizing every page in the index to help users with disambiguation. A search for “Apple” can be done in the category “technology” to avoid results about fruit. And when you search, SearchMe places the categories it thinks are important at the top. In the screenshot above, the query “safari” shows categories including “companies,” “software,” wildlife,” “photography,” etc. The image to the right shows category suggestions for “blackberry.”
Update: Louis Gray noted SearchMe was indexing his site back in February, and said “Is it taking a graphical snapshot, in the same way that www.archive.org has done to show how Web sites looked over time? I’m not exactly sure.”
Palo Alto based startup SearchMe has kept a low profile since being founded in March 2005. The company, which has 17 employees and raised $5 million from Sequoia Capital over two rounds, will launch a number of what founder Randy Adams calls “long tail search engines” in the near future. The first product they are launching is WikiSeek, which went live about an hour ago and will be officially announced on Wednesday.
WikiSeek is a search engine that has indexed only Wikipedia sites, plus sites that are linked to from Wikipedia. It serves two purposes. First, it is a much better Wikipedia search engine than the one on Wikipedia (and has been built with Wikipedia’s assistance and permission). Second, the fact that it also indexes sites that are linked to from Wikipedia means that, presumably, it will return only very high quality results and very little spam. It won’t show every relevant result to a query, but it will certainly give a good overview of a subject without all the mess.
The search results also include a tag cloud which contains Wikipedia categories containing the search term. Results can be quickly filtered by clicking on one of those categories (see screen shot, click for larger view). The first three results of a query are always Wikipedia content (unless there are not three results) and are shaded blue. The remaining results are below the shaded area.
In addition to the search engine, WikiSeek has two additional tools - a search plugin for FireFox, IE7 and Opera, and a really useful greasemonkey-like Firefox extension that will change the way Wikipedia looks on that browser by adding a “WikiSearch” button to the search box (see screen shot below). Click that button and see WikiSeek’s Wikipedia-only results. It’s faster and better than the results Wikipedia returns through its native search feature.
SearchMe is donating “the majority” of revenue generated from advertising on WikiSeek to the Wikimedia Foundation. Adams told me earlier this evening that WikiSearch is a showcase product for their technology, and they are happy to help the Wikipedia community as much as possible by donating those revenues.
Confusion with Wikiasari
WikiSeek will undoubtedly be confused with the much discussed Wikiasari search engine that was announced by Wikipedia founder Jimmy Wales last month. In fact, in our original post on Wikiasari, we included a screenshot that we later learned was not a prototype of Wikiasari. We corrected that post, and asked “the Wikisearch Screenshot Isn’t Wikiasari, So What Is It?” It was actually an early WikiSeek prototype, then called WikiSearch. Question answered.
