Hakia Promises Meaning-Based Search

It’s starting to look like the 1990s all over again, especially if you’re trying to keep up with all the new search engines trying to take on Google (or at least carve a niche out of the giant). So can we take it seriously when one of them promises something really different?

It might be better to ask whether we can afford to not take it seriously, at least for a first glance. After all, Google promised something very different with its PageRank concept and the idea that the measure of the importance of a web page is how many incoming links it garners. We can see where Google is now, but we can also see that web search still stands in need of improvement.

Enter hakia. The two-year-old search company recently received $11 million in funding from a range of investors, none of them "the usual suspects" from Silicon Valley’s Sand Hill Road. Hakia’s investors hail from all over the world. They seemingly have little in common, except perhaps an interest in high technology. They include:

  • Prokom Investments, based in Poland and active in the oil, real estate, IT, financial, and biotech sectors.
  • KVK, a Turkey-based distributor of mobile telecom services and products.
  • Murat Vargi, an angel investor from Turkey and a founding shareholder of Turkcell, a mobile operator and the only Turkish company on the New York Stock Exchange.
  • Lu Pat Ng, an angel investor in Malaysia.
  • Dr. Pennit Kouri, an economist and VC from Finland who apparently can’t stay still – he has taught at Stanford, Yale, New York University, and Helsinki University. He has also worked as an economist for the International Monetary Fund and sat on the board of Nokia in the 1980s.
  • Alexandra Investment Management, a U.S.-based investment advisory firm.
  • Bill Bradley, a name familiar to both U.S. basketball fans and political mavens, as he has played for the New York Knicks and served for 18 years as a U.S. Senator (D-NJ).

What is it about hakia that interests so many diverse entities? Part of it may be that Dr. Riza Berkan, founder and CEO of hakia, made the conscious decision to snub most venture capital companies out of fear that they would demand too much power. But there’s a lot more to the story than that.

Hakia was founded in 2004 by Dr. Riza Berkan, a nuclear scientist who worked for the U.S. government for 10 years. His major focus was on the safeguarding of information. Serving at the helm of a new search engine suits his specializations in artificial intelligence and fuzzy logic. Dr. Berkan was joined in founding hakia by Dr. Pentti Kouri (also mentioned as an investor), an economist and venture capitalist. Hakia also boasts a prominent scientific adviser: Victor Raskin, described by the company as "a father of ontological semantics and a noted international authority in the field of computational linguistics."

What this means, in short, is that there is a lot of brain power dedicated to making hakia work. In addition to the C-level names, there are 20 programmers working on the system in New York, and another 20 working remotely from various locations worldwide, including Turkey, Armenia, Russia, Germany, and Poland. All of these people are in search of one thing: meaning.

To be more specific, let’s look at hakia’s web site. On the "About Us" page the company explains that its "basic promise is to bring search results by meaning match – similar to the human brain’s cognitive skills – rather than by the mere occurrence (or popularity) of search terms. Hakia’s new technology is a radical departure from the conventional indexing approach, because indexing has severe limitations to handle full-scale semantic search."

So that implies a significant difference in hakia’s algorithm, to say the least. How do you get a machine to understand a complicated query when it’s tricky enough getting a search engine to tell the difference between bass fishing and bass guitars? Hakia uses a system it calls QDEX (stands for Query Detection and Extraction). QDEX analyzes web pages and breaks them down into something hakia calls "knowledge bits," which it stores as "gateways to all possible queries one can ask."

If that leaves you scratching your head a bit, don’t worry; it left me a bit puzzled too. What exactly is a "knowledge bit?" (I did in fact ask hakia’s search engine that question, with interesting results, which I’ll share later). Putting aside the software for the moment, hakia’s hardware includes "a distributed network of fast servers using a mosaic-like structure." The company isn’t too worried about the scalability of its system, either, because "data segments are independent of each other."

Hakia sounds impressive. It is trying to bring about search based on the actual meaning of content rather than keywords. But how well does it work?

It’s easy to find hakia’s web site. Just point your browser to http://www.hakia.com. You’ll see a web site with a very clean interface (for which we can thank Google as the trendsetter):

 

As you can see, it’s still in beta. Its blog has been around since February, in case you were wondering. It’s a nice place to check out if you’re curious about hakia’s special features; if you want to see them in a more organized fashion, click on About Us and then Benefits. Or you could just try out the examples:

 

Let’s take a look at short queries. It’s nice to be able to ask questions, but a lot of us are used to putting in keywords these days. So let’s ask about Orlando Bloom. What do you get? To the right of your results, you get a nice image that is linked to celebopedia.com. The results themselves are divided into "galleries." These have headings that include Biography and Timeline, Awards and Accomplishments, Filmography, Television and Radio, Speeches and Quotes, and much more. You get anywhere from three to five links under each gallery.

What if you have a more complicated question? How can you tell whether a link will be useful? Let’s try that Enron question.

 

Here you see two things. First, hakia does a good job of highlighting text that is relevant to your question – the whole text, not just your keywords. Second, the search engine makes a suggestion as to the most useful document, and lets you click on a link that takes you there quickly. Hakia calls this its virtual assistant (and I’d rather have that kind of virtual assistant than Microsoft’s paper clip any day). Sometimes the assistant seems to get involved in what you’re doing. When I used the sample question "how many people died in Iraq," it enthusiastically replied "You are working like a detective! Here is a good lead:…" and it pointed me to two links.

So, what happens when I ask hakia about its own technology? Well, I tried to ask it about knowledge bits. Here are the results I got:

 

Do you notice that each item includes "knowledge" and "bit," but not exactly in the way I intended? It looks as if hakia has fallen back on trying to find the keywords and matching them to the query. I want to emphasize that hakia is in beta, so I’m not going to take them to task for this.

I’d also like to point out that the way you phrase a question in this search engine makes a really big difference, even if the meaning is similar. The queries "What is the earliest known reference to juggling?" and "How old is juggling?" yield very different results. For the former, the very first result was relevant, but the rest of the results seemed to focus on the phrase "the earliest known reference to…" For the latter, hakia returned results pertaining to the age of the juggler, and "age-related juggling" (juggling glass eyes, bifocals, etc).

Hakia is working hard to promote itself to the right people, but so far it hasn’t had the effect for which it’s probably hoping. For example, when Kevin Maney at USA Today wrote about what he learned on the first day of the Web 2.0 conference, item three was "The bubble lives. You think there’s no Web 2.0 bubble? Yeah, well, explain an unknown search company, Hakia, sponsoring dinner and putting Hakia-signed white chocolates in the sorbets that were served after the salad. For a while, everyone thought the company was Ikea. What, a furniture search company?"

I also wonder about one of the recent releases from the company – hakia Search Music, "the Web’s first known search music CD." The creators ran random queries in hakia, found a search result they liked, and composed a song about it, using "ethnic instruments to highlight the richness and diversity of the Internet culture." I can’t decide whether this is a really brilliant or a really misguided way to promote the company and its mission.

Seriously, though, hakia does take an interesting approach to search. If the company can truly make its search engine understand content in a way similar to a human brain, it will definitely have something. Personally, I plan to check back with hakia a few times over the next month or so. I’ll run some queries to find out whether it can help me with my work. A search on "search engine optimization" led me to a list of useful-looking results and one of hakia’s galleries focused on search engines, so I’m quite hopeful. Time will tell how well it delivers.

Google+ Comments

Google+ Comments