The perfect search

John Battelle asks, what is perfect search?

Imagine the ability to ask any question and get not just an accurate answer, but your perfect answer -- an answer that suits the context and intent of your question, an answer that is informed by who you are and why you might be asking. The engine providing this answer is capable of incorporating all the world's knowledge to the task at hand -- be it captured in text, video, or audio. It's capable of discerning between straightforward requests -- who was the third president of the United States? -- and more nuanced ones -- under what circumstances did the third president of the United States foreswear his views on slavery?

This perfect search also has perfect recall -- it knows what you've seen, and can discern between a journey of discovery -- where you want to find something new -- and recovery -- where you want to find something you've seen before.

This is the Oracle of search. Not only would this search engine have to contain all the world's information, it would have to understand it and be able to determine the relevance of it to any particular task. It would have to build a model of each user, understanding what each person knows and doesn't know. Building this would mean we've built an omniscient perfect intelligence. Why this intelligence would bother spending its time answering our puny questions is a mystery to me.

However, perhaps we can take baby steps toward this goal. Much of the work in relevance ranking is going into further understanding the text of the page. Question answering systems tear through volumes of data frantically applying grammatical models and trying desperately to ferret out the answer buried in the sea of natural language goo. Personalized search, in some forms, builds a model of the user, understanding a little of what they know and don't know, and applies information gleaned from other similar users. We are getting better, but we're a long way from the Oracle.