Phase II Amount
$1,496,785
The last two decades have seen significant improvements to search and machine learning capabilities, but human-centered design of such systems has not kept pace. This is especially pronounced in domains where large data sets are handled, with applications built from a data-centric (how do we store this information), rather than user-centric (how is this information going to be used) standpoint. Conversely, the design of consumer software has made significant leaps forward, leading to a shift in user expectations that highlights the outmoded and archaic nature of the business tools.Interestingly, the convergence can also be seen from the opposite direction: where in the past, one might have dealt with dozens or hundreds of documents, its now common for a laptop to hold tens of thousands. The result is that even consumers need tools with a level of sophistication previously relegated to professional domains. We propose that solving such information problems requires addressing them from both sides: first, that using millions of documents should be as simple and fluid as using a mobile app, and second, that such an application should readily handle new document sets, making it feasible for everyday use on thousands of documents on an everyday laptop.
