In yesterday’s NY Times the article Cyberface: New Technology That Captures the Soul discusses the current most cutting edge approach to digital film-making. Up until now, motion capture was the cream of the crop, but this new company called Image Metrics has developed a method for directly translating a human actor into digital animation.

What is supposedly so groundbreaking about this is the level of human-ness it is able to convey through the digital animation (as implied by the title’s ‘capturing of the soul’). But what I find interesting is the concept the creators built this new approach on. Here is the quote, I’ll discuss after:

Many surveillance devices rely on facial recognition software, but it produces a lot of false positives. Mr. Edwards and his colleagues took a different approach, one that starts with the generic model of a human head and layers onto that a mathematical distillation of an individual’s expressions. He compared his approach to describing a new bicycle. The person who’s listening is likely to picture the new bicycle based on other bicycles she has already seen.

“It’s model-based computer vision,” Mr. Edwards said. “The idea is, if you know an object, you can picture it. The key for animation was that realization: that we needed to build a computer system with the prior concept. The mathematical structure describes the basic concept of the face and maps the subtle variations.”

I find this interesting because, essentially, they looked into the human mind for clues on how to improve this technology. After analyzing the way our minds work when, say, picturing a bicycle that is being described to us, they culled out a ‘truth’ about the way our minds work, and applied this to their software.

This doesn’t give me any answers at the moment… but it does give me a question to ask: What methods do human beings use in order to determine the level of relevance of a particular document? Can an in-depth analysis of how humans perform this function, re-inform the path to higher accuracy in the search engine world? Can any new starting points be discovered or uncovered that have yet to be?