I agree with this need, and a certain level of user will find this very helpful.  However,  there is a larger group of users who will be looking for graphical displays (charts, graphic representations of the numbers, etc.).  For those users, drill down capability embedded in charts and graphics is very helpful and intuitive.  So for instance,  if they click on a line on a chart it opens another window that expands to shows a graph of that particular piece of date.  Drill downs could go all the way down to individual project, contractors or subcontractors.  Having both types of data analysis and presentation available would reach a broad audience.

The idea of semantic lenses is all about having a flexible way to present recovery.gov data. Suppose recovery.gov collects data on an (imagined) recovery project involving restoration efforts in our National Parks.

With a traditional approach, information about the money spent on this project would be made available via a user interface designed by recovery.gov itself. Maybe there would be a table showing how much money each National Park was getting.

Now suppose a visitor wants to instead plot this information on a map, or see a view that mashes up the National Parks spending with other stimulus funds spent by the Department of the Interior. The idea of semantic lenses is that any visitor to the recovery.gov can choose to create a new map view, and then (simply by clicking!) choose what data to put on the map.

Think of it as "choose your own interface".

Regarding implementation approach, XML is a useful interchange format for representing data, but it does not capture the concepts in the data that are needed to enable this sort of scenario in which users can mash up data & views however they want. A semantic (RDF-based) approach provides the conceptual type information needed to flexibly combine data with user interfaces (into "lenses") in an as-needed basis by end users of the recovery.gov Web site.

Both the semantic lens and drill down approaches call attention to the need for information to be readily visualized and navigated by diverse sets of users.  I would add that these tools could be used for additional, and much broader, purposes - especially around the widespread cynicism and distrust of government that exists in the U.S.

In addition to these valuable technical tools, I urge you to take into consideration the "moral semantics" (motivational values and priorities) of users can be influenced by the Recovery.org platform.  Of particular concern is the lack of public confidence in government.  Recovery.org could be used as a platform for increasing public confidence, but only if it takes into account the existing motivations for distrusting the government and address them.

This deeper level of semantics is at the heart of politics, and should be considered as part of the technological platform of this project (treating moral semantics as an applicable body of knowledge is akin to it being a social technology).

I agree that adding semantics is essential for making sense of the breadth and depth of data available, for discovering patterns of use and potentially abuse, and for supporting a variety of useful presentations based on citizen interests and areas of expertise.  I also think that some novel approaches to navigation and visualization will complement the use of semantics to make the more widely accessible, and thus open the door for greater citizen understanding and participation.