VentureBeat

TextDigger, a semantic search startup that launched early last year at DEMO, has been much quieter in the interim than other companies in the space like Hakia, Powerset and Radar Networks / Twine. But now the company has come back to light, at least for us — a filing document reveals that the company has finally landed venture funding, raising $3.8 million so far.

A year or two ago, it wasn’t uncommon to hear the term “Google killer” applied to companies like TextDigger. Expectations have since fallen somewhat, and most of the semantic startups have struck out in directions other than plain-vanilla web search.

TextDigger landed on media services — no surprise, given that it was founded by former CNET employees. The company tells me it’s making the most progress with automated tagging and keyword generation, which help websites become more structured and show up higher on search rankings. Another product, a related search technology that helps visitors navigate a site, has also gotten interest.

The products are completely developed, they say, and open to any publisher that wants to sign up. TextDigger makes money by charging for all three offerings, following a trial period; related search is charged per 1,000 queries, while auto-tagging and keyword generation are charged per query or tag. It will add self-serve tools this summer to help streamline and automate the process of signing up for the services.

It’s encouraging to see TextDigger making some progress, and figuring out how semantic tech can feed hungry (investor) mouths. But notably, it hasn’t entirely given up on the idea of web search. Although the search segment of their website is closed to the public, when I asked whether the company had given up, a company exec responded that working on search and navigation on other sites would ultimately improve their web search, writing, “If and when we launch a public Web search, at a later date, it will certainly be much better for having incorporated the fruits of our labors in these other service areas.”

By contrast, Hakia, which launched around the same time (our first coverage on the two was in the same article), has continued to make web search the core of its business, with new offerings like vertical search technology.

However, Hakia is also looking to other businesses for its revenue right now; earlier this year, it began licensing itself for enterprise search applications. And today, it’s announcing a new push to syndicate its engine onto other web sites, along with some bells and whistles like XML feeds. It’s offering up to 30,000 searches per day through a web site for free, before it begins charging.

It should be interesting to see how these two companies progress compared to each other, since they appeared at nearly the same time, with a similar pitch, but have since taken on very different strategies.

TextDigger’s $3.8 million round was led by True Ventures, with participation from the Exis Trust and CNET, which provided the San Jose, Calif. company’s $1.5 million seed round, and also uses the technology on its own website.

Semantic search engine Hakia is following up its recent licensing news with a new scheme to attract users: Results that are narrowed according to the vertical you’re searching, starting with healthcare.

When you search on a health-related subject in Hakia — whether you’re looking for a medication, treatment or condition — the engine will recognize it as medically related and restrict its initial search to “credible” resources, in this case, those vetted by the Medical Library Association.

Hakia’s engine has been slowly improving over the last few months, to the point that most searches return a mix of useful and not-so-great results, which is about as good as Google has ever done. Restricting results to select sites is a way for the company to seem to accelerate development without more heavy-duty work on its core technology — essentially, a clever trick.

Trick or not, Hakia has landed on a way to differentiate itself and try to capture a few more users. It’s an interesting approach; Hakia could potentially limit results according to any number of verticals, such as travel, sports or finance. The ability to recognize queries related to a particular subject is pretty neat and helps to filter out some of the junk included in regular searches.

That doesn’t mean the company is following Powerset’s example of working only within certain domains (mostly Wikipedia, with some Freebase thrown in) to improve the results returned from those specific sites; Hakia is just performing a regular search on its limited selection of healthcare sites. Hakia’s suggested search for the question “What are the benefits of aspirin” (try it out here) is far from perfect. A simple search for info on prilosec, shown below, was more impressive.

Powerset, which I also recently wrote about, does want its results to be perfect (as much as possible), although that will limit the speed of its expansion. Hakia, though, seems happy to try to exceed Google’s usefulness by a thin margin and hopefully win over followers with time as it continues to improve its index and technology.

I tried out a few other searches, and got mixed results — take a look here and here. Readers, how valuable do you think Hakia’s technology is at this point? Is vertical-based search an improvement?

Semantic search engine Hakia has started to license its technology to other startups, starting with RiverGlass, a company that digs through and summarizes information on behalf of government agencies.

Semantic search technology seeks to understand word and sentence meanings in order to provide you better results than the key-word approach used by say, Google. Semantic search competitors such as Powerset hoped to become a destination search site that would compete with or at least complement results provided by Google. But getting semantic search technology to rival Google’s results has been slow going.

The four-year-old Hakia has accumulated some $21 million in funding to support its development, and has grown to about 50 employees. Like other companies working with semantic technology — Powerset and Radar Networks being two — it was branded initially by the media as a potential “Google Killer”.

Last year, Hakia launched Hakia Challenge, a side-by-side matchup of Hakia and Google. The results sometimes compared well to Google, but generally did pretty badly. At this point, the results are better. At the least, Hakia provides a usefully different perspective from Google’s results (there’s an example after the break). But Tim McGuinness, the company’s VP of search, says that the engine never intended to compete with Google. The side-by-side comparison was merely done because so many people requested it.

But the company always planned to build its technology for specific applications, rather than offer the full-web search Google does, he said. Hakia is turning to license its technology, because it can helpful for various tasks.

These tasks include information analysis, summarization of documents, machine translation and terminology standardization (in which similar words like “separation” and “divorce” would be changed to a common term, because in some instances users don’t necessarily want to differentiate between the two). The process is supported by Hakia’s internal reference table, which includes over 100,000 words. Any task automating the processing of data can benefit from having Hakia involved.

Hakia is actually closer to natural language processing, in which computers try to understand a sentence by its structure and meaning the way a human would, than it is to semantic search, which can be done in various ways but generally means finding interrelations between terms in bodies of text.

McGuinnes says he doesn’t think the company has any direct competitors who are trying to do the same thing. While that may be true, other companies like Powerset are also building out APIs for developers to plug into, and regardless of the methods used, the companies with the most accurate technology in any given situation will win out, as occurred with Google in web search.

Update: Hakia’s CEO has posted a response on the company blog to clarify that, well, they’re happy competing with Google too, if that’s an option. “hakia’s competitive position is undefined, and hakia’s promise is not built on competing for the same turf with others … Semantic technologies will bring out something new about the Web that is hard to place in any competitive scale,” he says. Read the rest here.

Read the rest of this entry »

Hakia, another search engine company that is trying to understand “meaning” of your search terms, as a way of competing against Google, has raised $5 million more in capital from Prokom Investments, an existing investor.

The investment comes at a time when activity in new search engine approaches is robust. Just today, Wikipedia’s Jimmy Wales launched a new search engine, Search Wikia, that hopes to recruit people to edit its results. Mahalo is another people-based search engine, and Silicon Valley’s techies have more in the works.

However, Hakia, based in New York, has gotten less buzz than the others. It is far away from the cluster of scientists and engineers here in Silicon Valley that have helped push the search engine industry forward, and its investors are unconventional. Prokom, the latest investor is a Polish private equity firm, which besides making tech investments, also back biotechnology, real estate, construction and oil companies.

Hakia’s other backer are private equity firms too. Private equity firms typically invest in later-stage companies, not early ones, like Hakia, which are still in the process of putting their team together and in the early stages of figuring out their technology.

The latest funding brings total investment to $21 million. The company says it is now building a semantic advertising platform and a European datacenter. Other investors include Noble Grossart Investments Limited, a subsidiary of Noble Grossart Limited; Alexandra Investment Management, an investment adviser; and KVK, a large mobile telecommunications company in Turkey. Individual investors include Dr. Pentti Kouri, who is also a co-founder of the company; Lu Pat Ng, representing his family from Malaysia; and Murat Vargi, an early investor in Turkcell, a leading GSM operator of Turkey.

Here’s the latest action:
1. Medio’s roar turning to a squeal?
2. Sprint changing WiMax plans?
3. AOL rumored to be considering buying ad targeting network Quigo for $300 million
4. Bug Labs, for open-sourced electronic devices
5. Semantic search engine Hakia releases social networking tool
6. The amazing $200 Ubuntu Linux “green” PC at Wal-Mart
7. Cisco does its 125th buyout
8. Facebook’s stock has appreciated 33-fold, and then some
9. Internet Brands going public with growing losses, declining sales?
10. Shopstyle signs deal with In Style Magazine

Medio’s roar turning to a squeal? — We’re wondering what will happen to Seattle’s Medio, the company that provides mobile search technology to telecom giant Verizon, now that Google is reportedly close to signing a deal with the giant carrier to offer customers a GPhone. This which would carry a Google operating system, based on Linux and offering a host of Google applications including search. The Mercury News carried a profile story on Medio Tuesday, in which chief executive Brian Lent (pictured top left) boasts Medio has a broader reach than Google, citing its partnership with Verizon. The article appeared the day before the deal negotiations between Google and Verizon leaked. Obviously, Medio won’t get kicked off of Verizon’s phones overnight. The GPhone hasn’t even arrived yet. Still, the Merc piece is a notable read, explaining how Lent knew the Google co-founders at Stanford, but initially shrugged of the promise of search; he even turned down the No. 1 employee position at Yahoo. Verizon, in turn, may be flirting with Google for the following obvious: Apple iPhone has become a raging success, after Verizon turned down the chance to be the iPhone carrier partner. As Techdirt points out, Verizon and Google have a tough history, including the standoff over the 700 MHz spectrum debate and network neutrality. But the GPhone, that might paper over the differences.

Sprint Nextel changing WiMax plans? — Sprint may be rethinking its plans to offer high-speed wireless Internet service using WiMax technology, possibly merging its wireless broadband unit with start-up Clearwire, according to the WSJ.

AOL rumored to be considering buying ad targeting network Quigo for $300 million – But its just a rumor. Details from Kara Swisher.

Bug Labs, for open-sourced electronic devices — You’ve heard about all the open-source software. Well, New York’s Bug Labs is offering open source for hardware, drawing on outside developers to help fashion the building blocks of these personalized devices that will be easy enough for non-techies to assemble. It is backed by Spark Capital, Union Square Ventures and Robert Young, founder of open source company Red Hat. More about BugLabs here. It has similarities with Ponoko.

Semantic search engine Hakia releases social networking tool — Called Meet Others, it lets you meet people who typed in the same search query. Now that is geeky.

The amazing $200 Ubuntu Linux “green” PC at Wal-Mart — The price of this computer is truly in the basement. It runs OpenOffice software and comes pre-configured with links to all of Google’s online applications. See Wired story for more details. It stands in stark contrast to Nicholas Negroponte’s $100 laptop for the poor, began at a price of $100, but which now has crept upward to….$200. So apparently, there’s no need for the poor laptop any more. The developing world may as well order a Linux version straight from Wal-Mart.

Cisco does its 125th buyout –Cisco, the giant networking company, proves you can grow and prosper through non-stop acquisitions. It has paid $100M for Securent, a company that monitors access to a company’s data and communications regardless of vendor, platform, or operating system. Securent was founded in 2004 by Rajiv Gupta after raising capital from Greylock and Onset (via Alarmclock ).

Facebook’s stock has appreciated 33-fold, and then some – Many of us reported how the Microsoft deal to invest in Facebook at such a high value ($15 billion) makes Facebook’s private shares expensive. This, in turn, makes it tough to recruit motivated employees, because it means there’s little room for the stock to appreciate — at least for several years until Facebook starts making some money. NYT’s Miguel Helft has done a good job at exploring just how far the stock has risen.

Internet Brands going public with growing losses, declining sales? — We don’t get this one. The El Segundo, Calif., company has filed to raise $45 million in an IPO. But the company, whose sites include CarsDirect, WikiTravel, FlyerTalk.com among others, swung to a loss of $2.4 million in the first nine months of the year, and saw is revenues decline too. This, after buying 35 start-ups last year, enough to jolt any company. Buyer beware. It is backed by IdeaLab.

Shopstyle signs deal with In Style Magazine — ShopStyle, the fashion-focused shopping search engine that Sugar recently acquired, has announced a partnership with the popular In Style Magazine. Sugar has said its efforts to create a woman-centric network, a la Glam, was not working out, and has been looking into other means to generate revenue. Combining ShopStyle’s search with the editorial sensibilities of In Style’s fashionistas could prove lucrative for both companies. Under the terms of the deal, Sugar will receive a cut of revenues, as well as its CPCs.

Here’s the latest action:

Murdoch appears to have enough Bancroft family support for purchase –The jury is still officially out, but the NYT is saying it looks like Dow Jones and its jewel, the Wall Street Journal, will indeed to go to Murdoch’s News Corp.

Hakia’s Scoop Bar — Hakia, one of the new search engines trying to take on Google by using “semantic” technology, has released a so-called Scoop Bar, which takes you more directly to the text you’re looking for in results. When you select a result at Google, you get taken the page and that’s all. Hakia’s feature automatically scrolls down to where the good stuff is and highlights it. (Via Pandia).

Hitachi’s new search — The company has developed a search technology that can find images similar to a specified image from millions of images and video data in one second.

Wikia buys Grub — Wikia, the company that wants to draw on user participation to create a search engine that rivals Google, has taken another step. It is buying Grub, an screensaver users can download and which exploits their idle computer CPU time to crawl the Web. Search Engine Watch has a good story on this, explaining how the Grub open source technology works. GigaOm has an informative post too.

Yahoo Search Assist — These are the days of new search engine features. Yahoo now has a feature called Search Assist, which seeks to detect when you are hesitating about what search term to use, and then offers up alternative queries. It goes beyond guessing what word you’re typing. It offers up terms that are also conceptually related to the search query you’re using. There’s a good review at SearchEngineLand.

Live Search’s new commands for search — Microsoft has unveiled new short commands for image search. If you’re looking for images of Jimmy Hendrix for example, you can add “filter:face” to select only images of Hendrix’ face. There’s also “filter:bw” for black and white images, and “filter:portrait.”

Google enterprise — This software lets companies and other groups view their global data and imagery. Companies are using it, Google says, for things like designing new buildings, exploring for energy and responding to emergencies.

Cisco may have acquired Click.TV? — Cisco may have acquired a video annotation and deep tagging service called Click.TV which shut down last month, according to Techcrunch. A Cisco spokesman responded: We do not comment on rumors.

Google is promoting its Checkout more aggressively on its site — Details here.

PG&E to buy electricity from one of the world’s largest solar plants — Enough to power 400,000 homes, the power comes from a deal with Solel Solar Systems of Israel, which is building a 553-megawatt plant in California’s Mojave desert.

TextDigger is the latest company seeking that Holy Grail: Improving on Google’s results by understanding the sense of the words you’re looking for.

TextDigger’s search engine is called Digger, and it just launched at the DEMO conference.

First, some context: Digger, of San Jose, joins Powerset, the San Francisco start-up, and Hakia, of New York, and others which are trying to do something similar. Our piece on Powerset sparked debate within the search industry, namely because some experts believe automated search engines can never understand words the same way we humans do. Google, many think, is doing about as good a job as is possible. Google doesn’t try to assign meaning to your search term. It merely ranks the pages it thinks are the most popular that contain those words. Powerset still hasn’t launched. Hakia has, sort of, which we’ll get to in a minute.

TextDigger addresses the semantics problem by asking users to help it. If you search for “hotel with a view of the golden gate bridge,” it will tell you what it assumes about your intended meaning, and then asks you to modify. “View,” it tells you, means panoramic view, as opposed to personal belief. That might be fine, but what if you want personal belief?

By clicking on “refine keywords,” you can tell it you want “view” to mean personal belief (see screenshot below), or both panoramic view and personal belief. In other words, it is a classic Web 2.0 company, getting the masses to work on its behalf, and in turn improve results for everyone. This is significant, because Digger launches ahead of Wikipedia founder Jimmy Wales’ effort to do something similar. Digger relies on what people tell it about semantics. Powerset, by contrast, relies on smart grammar interpretation, which is very different.

Whether Textdigger can pull this off is another question.

It is in a closed testing period. You need an invite. Chief executive Tim Musgrove says it gets better the more people use it. It’s not ready for the masses.

We played with Digger. Our conclusion: We think this is useful tool, and we’d be surprised if Google didn’t implement something like this soon — especially if Digger gets any traction. It should be easy to do.

We should note, with Digger, you can choose to refine the word meaning for only yourself, or for the community at large. And there are shortcomings. Digger dissects meanings of individual words only, so it can’t assess the meaning of two or three words together, like Powerset is trying to do. The other shortcoming is that depends on you logging on at the site, to personalize results for you. It will have a hard time opening up to non-registered access.

TextDigger has received $1.5 million, led by CNET, where Musgrove was a researcher.

.

Meanwhile, Hakia has quietly launched.

Check out what it is doing with searches for people, like for Madonna. It categorizes results into news, music profile, biography and more, rejecting the repetitive results that you get at Google. Similarly, look results for chocolate, India, The Beatles or Red Sox. It’s too early to tell what exactly Hakia is doing, as we’ve yet to talk with them. But it’s essentially a hybrid of Google and Wikipedia — a more aggressive push toward the categorization strategy that Ask has been heading toward (see the right hand side of this search for chocolate, for example).

Round-up of the latest action in Silicon Valley:

It’s all about Google — Notable post by Rich Skrenta about how Google is becoming ever more important. Having become the “start page” for the Web, Google makes $90 to $120 for every thousand times its pages get viewed (or CPM), compared to a mere $4-5 for an average page view elsewhere, he calculates. We asked him for his assumptions, and it turns out he mixes a lot of sources, but the high-end of his figure comes from an analysis from Caris & Co.’s Tim Boyd, who uses Comscore data to estimate that Google makes between 19¢ and 21¢ per search (or roughly $200 per thousand searches).

Google boosts its own products in search results — Google continues to find ways to upgrade its own features in search results. Blake Ross has a good piece about the latest, “Google Tips,” which lists Google’s photo-sharing site, Picasa, a tip for users at the top of its results, even though Picasa doesn’ even show up on the front page of its own search results (see image below). That’s just one example among many such self-interested tips. We messaged Google last week to see when they introduced the feature; they didn’t respond.

Too many others chasing Google — Minekey is just the latest search service to appear, with $3 million in funding (see our story here). The NYT follows today with a piece about all the other search engines vying for a slice of the all-pervasive search engine market. We’ve covered most already at VentureBeat. Problem is, their logic doesn’t quite hold. Yes, if you pick up a single percentage of the search market, you’ll get profits, but it no one seems to be pulling this off (explained by Skrenta’s “winner-take-all” analysis above). Among others, the NYT piece mentions ChaCha, the search engine that has live guides help you with search (we’ve used it, and it’s true) has raised $6 million, including from Amazon’s Jeff Bezos. Our previous coverage of ChaCha is here (scroll down). It’s hard to comprehend, though, how the company will cover labor costs. The NYT also mentions the yet-to-launch Hakia, a company reported on elsewhere, but not yet mentioned by VentureBeat — which wants to do something similar to Powerset also promises: a natural language search. Hakia raised $16 million from Noble Grossart Investments, Alexandra Investment Management and a motley group of others, including Polish and Turkish oil, real estate and telecom groups.

WiFi in cars; watch out for accidents later this year — Autonet Mobile, a San Francisco wireless start-up, is expected to announce this week it has partnered with Avis Rent A Car System to “provide a rolling WiFi hotspot to Avis customers by March,” according to the NYT. For $10.95 a day, Avis will give customers a portable device that plugs into a car’s power supply and delivers a high-speed Internet connection. It will use the 3G cellular network and work in all metropolitan areas and “about 95 percent of the country,” says Sterling Pratz, the company’s chief exec. We’ll believe this when we see it, because many parts of the Bay Area have spotty cell network coverage as it is, let alone 3G coverage. There’s an amusing concession in the story. The company has some technology limitations, the Times says, like “bandwidth restrictions.” Meanwhile, Ford is introducing bluetooth systems next year in its cars, allowing people to chat, and check email.

Gmail vulnerable to contact list hijacking – If you visit the wrong Web site, you may be vulnerable to having your Gmail contacts lifted. That’s because Gmail stores its contacts in javascript files. Details here.

Walloped takes a wallop — Sean Uberoi Kelly, CTO, founder and principal inventor of Wallop has left the San Francisco social networking start-up Wallop, according to Gigaom. Chief exec Karl Jacobs says the departure is normal, noting Kelly is going to back to New York, where he has family — and has finished building the product. Fair enough, but it’s a blow nonetheless. We have yet to understand Wallop’s strategy and business model (explained here in detail). It raised a total of $13.6 million from Norwest, Bay Partners and angels.

Gov. Arnold Schwarzenegger’s investment plan for California tech — A hodge-podge of projects comprise the governor’s $95 million proposal for research spending. Summary in Merc: Some $30 million would go toward building a new research building for the Lawrence Berkeley National Laboratory’s Helios Project, which is developing the next generation of efficient solar-energy technology. An additional $40 million would help build an Energy Biosciences Institute facility for research into alternative fuels, at either the University of California-Berkeley or UC-San Diego. Some $5 million would help the University of California’s bid to build a $200 million super-fast “petascale’” computer at the Lawrence Laboratory.