Trovix and the Next Generation of Search Engines
Reading all the techie articles about semantic search, statistical models and the semantic web, it occurred to me that in one way, all that debate is besides the point. The truth is, consumers don’t really care about the technology under the hood. They just want what it is they’re looking for.
The problem is, people are terrible when it comes to telling computers what they’re looking for. They use just two or three words to describe what they want. They’ll enter “Vacation, Hawaii” or “Pain, Knee” and expect a search engine to show them . . .well, show them something.
On job sites, the average search is 2.2 words and a location. What’s the right job for someone who wants “Full-time, California?”
With only two words to go on, the best search in the world is still going to have a hard time understanding what the searcher wants. That’s true if you have semantic search, statistical models, or the Head Librarian from the Library of Congress trying to get the “best” answer.
I think real innovation in search is going to have to come from creating user interfaces that get consumers to form better search queries. For web search, we’ve had the single search box for more than a decade. For travel sites, it’s still two cities and two dates. For most job sites, it’s title/location. We’ve trained people that searching should be fast and easy, and then lowered their expectations about what results to expect.
Dating sites figured out years ago that to get a person matched up with the best result, you need a lot of information. The more important the answer is, the more time someone will invest in the search. eHarmony figures that finding the love of your life is worth a half hour. Google figures that finding pictures of Britney Spears is worth .09 seconds.
What about searches that fall between the extremes? I’m always frustrated when searching for cars and vacations because you have to know exactly what you want. You can search for Jaguar or Austin Martin. But you can’t search for “off beat roadster.” With vacations, you start by picking a departure date and destination. I’m flexible on both. I’m happy to surf in Costa Rica, Mexico, or Hawaii, depending on the cost of the hotel and availability of flights. A travel agent can do this for me, but the travel sites can’t.
The next generation of search tools needs to take the front end challenge as seriously as it takes improving the back end technology. Better search results are going to have to come in large part from allowing people to run more human-like searches.
Trovix has a unique approach for job search. Our interface requires people to upload a resume. (Something most professionals and job searches have already.) We use the entire document to create a semantic model of the person’s skills, experience and likely career interests. Just like a person can look at your resume, guess what jobs might be good for you, we can automatically match you to jobs based on your past work experience. We combine that with explicit search terms to create a search that is personalized and truly accurate.
The process takes less than 30 seconds, so it’s not a huge investment from the user. And the time they spend uploading their resume, they get back in spades by not sorting through dozens of bad job listings.
As everyone is looking forward at how the next great search innovations are going to change the web, it’s worth thinking about how these innovations will impact the general search user. With two and three word searches, the quality of the engine won’t be the limiting factor in search. The one-size-fits-all search box may be the biggest thing we need to innovate away from.
Thanks to Daniel Enthoven. If you want to know my thoughts on Next-Gen Search, you just read it.
