OSINT, Google and the USPTO

So we are familiar with the USPTO...
Traditionally searching it is a nightmare since the interface is so bad and no real effort to facilitate searching has been done.
So along comes Google and creates the mostly wonderful patent search engine. A sigh of relief? Somewhat but not really. For some reason they have decided not to do the most obvious thing: create a really good solution to this mostly not very difficult problem.
Patents are a quite limited and structured form of data. They are limited in numbers, are mostly text, have set internal structure and self reference already built in. This would be a classic case where OSINT with its various application of data mining and classification combined with stronger semantic engine and new forms of graphic representations.
Some examples of what I am talking about can be found in the the ever changing space of specialized search engines that pop up in the web2.0 sphere. Take a look at Mooter and Axioma for a couple of examples.
So what do I propose? Here are some ideas that come to mind (not exhaustive by any means):

• Use tags, labels and color coding for critical information (like author, assignees etc.)
• Create visualizations of datasets (e.g. all patents by author and what fields they are in, what space is a certain company involved in)
• highlight interconnects (who cites this patent? what field are they applying this in?)
• "People reading this patent also read..."

Of course putting in the provisionals is also key...and maybe integrating PCT...

I am admittedly not familiar with professional grade patent search tools but from what I hear they also lack these features and depend on hard work, deep knowledge of the database and material and luck...

Why has this not happened?

MC