Untitled

I think a lot of the early, and current, search engines have mis-managed expectations of what they can deliver. They’ve offered a simple search box with a bit of instruction and imply that’s mainly what the user needs to find what they’re looking for.

Imagine if you walked into a library and had a question but didn’t know what kind of collection to start searching through. You might seek out a librarian to help you. Would you approach them and say “London weather information”? Or would you say something like, “I’ve just started a new job which will require me to travel to London at various times throughout the year and I’d like to find out what the weather is like at all of those times.” The former is what we type into a search engine query box, and the latter is how much information the search engine actually needs to give you decent results.

A site like Ask Jeeves has done a bit to correct those expectations. They’ve changed three things:

They not only accept but encourage natural language queries

By accepting natural language, a user is more apt to phrase a question more thoroughly (not scientifically proven, just my own theory)

They index documents by subject matter and not by word content and frequency

Other ways of doing this:

Encourage, by examples for example, the writing of longer queries

Use a multi-step process, a wizard perhaps, to zero in on a result.