Update: If you’re looking for deep dish pizza near O’Hare, see my step-by-step guide to Gino’s East on Higgins.

Having gone to school in Chicago, I love deep dish pizza. Unfortunately, there’s no Carmen’s or Giordano’s in the D.C. area. The last time I had good Chicago-style pizza was when my friend Jason flew in a few Giordano’s pies for his Super Bowl party. (The Colts were are also represented with tenderloins.)

I was connecting through O’Hare today and wanted to get some deep dish at the airport. I asked Google for “deep dish pizza at o’hare”. No luck.

This is a really difficult query for search engines. It seems simple, but it has a lot of components that make it tricky. But it’s exactly the kind of query that search engines should be able to handle.

Breaking apart the components of the query, we have:

“deep dish pizza” is a distinct concept. It’s different from “New York pizza,” “Sicilian pizza,” and “Indiana pizza”. (I don’t know what that is, but my friend Wanita swears there’s such a thing.) I could restrict my query using quotation marks around the phrase “deep dish pizza” but I shouldn’t have to do that. On the other hand, “deep dish pizza” is close enough to “Chicago-style pizza” that those results should be included.

The second part of my query was “at”. Search engines typically treat words like “at” “and” “near” and “or” either as filler and ignore them, or they use them as Boolean operators. There’s a big difference between the query “deep dish pizza at o’hare” and “deep dish pizza near o’hare”. With 90 minutes between flights, “near” doesn’t work.

“O’Hare” is also tricky. It’s a known place with a physical address. But Google and other search engines know it as ORD or 10000 Bessie Coleman Dr, Chicago, IL 60666. Compare the results for “deep dish pizza o’hare” with those for “deep dish pizza ORD“. Frequent travelers might shortcut to ORD, but again, that’s not a burden users should have to bear.

The answer, in theory, lies in natural language search. I’ve written before about how search engines force people to think like computers. Natural language search tries to teach computers to think like people. The most talked about company in the space is Powerset. I saw a controlled demonstration of their technology in August, but the promised fall public beta has yet to materialize.

Keyword-based search engines fake some of this by using tricks like stemming, synonyms and anchor text. With the uptake of sites like Yahoo! Answers and the sheer volume of information on the Web, there’s a decent chance that someone has phrased the question the same way. In the search results page for my original query, one of the results was a Frommer’s Q&A.

In addition to the structural challenges of queries like this, there’s also the challenge of how data is gathered. Data providers do a terrible job of gathering information about a place that’s really a collection of places — such as malls and airports. In some cases, information is simply not collected. In others, the information that is collected isn’t sufficiently descriptive. The physical addresses of these businesses aren’t meaningful to users. “Terminal 1, Gate C3” makes sense; 10000 Bessie Coleman Dr, Chicago, IL 60666 does not.

OK, how many geeks are pulling out their laptops and doing searches like this you ask? Not a lot. And in search from the Web, it’s relatively easy to re-do the query and keep tweaking it until you get an answer.

Getting better answers faster becomes increasingly important as search moves to mobile devices and with voice-based search from the likes of Tellme and Google’s GOOG-411. In those environments, the penalty for failure is much higher. Users can’t easily tweak queries. They can’t browse endless Web sites to try to get the answer. They need the algorithms to do the work for them.

I was finally able to find out about pizza options at O’Hare by going to the O’Hare Web site and looking at a PDF map of Terminal 1. There isn’t a deep dish pizza place in Terminal 1, though there are Pizzeria Unos in other terminals.

The pyschic search engine would know that Pizzeria Uno is not an answer that works for me.

More on: local search, search, wireless