Better Search Through Fuzzy Results

I had a nice chat this morning with Steven Lavine, CEO of Transparensee Systems, about his company’s new offering, the Discover Search Engine (DSE).

DSE uses fuzzy matching to analyze database content and structure, so that it can return results that match a user’s intent… even without matching the exact terms specified by the user. It does this by offering a parameterized query interface; what’s considered a match and how important any given parameter is can be controlled on a per-parameter basis.

Example’s always help in groking these sorts of concepts, so here’s one to work with…

You click over to your favorite restaurant review site to plan a date. You’re looking for a nice place for dinner near the theatre you’ll be seeing a performance at later the same evening. Using a DSE driven interface, you specify parameters for: food quality, service, decor and price, select a food category (e.g., Greek) and specify the town the theatre is in (e.g., Springfield).

As luck would have it, there are exactly zero matches for a restaurant that exactly meets all of your criteria (which is exactly what most search engines would return). That’s where DSE shows value; instead of returning an empty result set, it gives you results that are as close to what you’ve specified as possible, ranked by the parameters of greatest importance.

For example (assuming price was most important):

– A restaurant that matches everything you’re looking for, but is in a town adjacent to Springfield shows first; followed by…

– A restuarant that is a 20% more expensive, but matches all of your other critera; followed by

– An Italian restaurant this is otherwise a perfect match; etc…

The bigger picture here is, for most of the searches we currently do, there is more than one acceptable ‘solution’. Fuzzy search through Transparensee’s DSE surfaces these ‘substitutes’, while ensuring that they are still relevant by controlling acceptable bounds for each of the parameters searched on (e.g., 10% more expensive is ok, but 50% is not).

Applications for this type of technology are as broad as your imagination. Some obvious ones include: dating, retail/product catalogs, real estate, and local events.

Transparensee is currently licensing their product to a number of small start-ups, and is in active talks with a few top-tier category players. Like most early stage companies, they are actively working on their pricing model and sales strategy, and are therefore offing classic enterprise + maintenance, per-user and ASP licensing/pricing.

While some companies have built similar internal proprietary solutions, I suspect there’s a large (and underserved) market amongst traditional publisher’s/media co’s who would be better off licensing a solution like DSE vs. building and maintaining it themselves.

Posted in Uncategorized