The race is on to make (and sell to users) the first truly good search engine that deals well with concepts, such that a search for “dog bites man” would include results without the word “dog” or “bites” but that include “Pomeranian” and “attacks,” to give a very simple example. Natural language processing — or NLP — is not easy for machines to learn, of course. Not only must they have a decent thesaurus, but as well they should be able to parse a document and derive some sense of context so that the results of a “dog bites man treatment” search might differ appropriately from a “dog bites man penalty” search. And so on.
Cognition is one of the companies boasting a pretty effective NLP search technology. (Others worth looking at, if the topic interests you, are Powerset, [true knowledge] and Hakia.) What makes Cognition particularly interesting to us is that it has focused its efforts in the two specialized markets of law and medicine. LexisNexis uses Cognition technology within its Concordance case management product, and Cognition offers it to individual law firms as well.
You can give it a spin to see how it compares with a straight-up Google search or your best Boolean concoction: CaselawCognition searches a database containing U.S. federal courts of appeal and Supreme Court decisions [1950 to the present: PDF] made available online by Public.Resource.org. It might even save you a nickel or two by finding that pesky U.S. precedent quicker and for free.
Oh, and I have to say that a search for “man bites dog” didn’t turn up anything really astonishing, though Rotweiler did figure once in the 19 results. Mind you, I could have shifted things, I suppose, if I’d played with one of the drop-down “meaning menus” for the words I used. As it was, “(domestic) canine” “male person or generic human” and “sink teeth into” were pretty much on point.