agency name standardization, notability

[John Wonderlich]

Two quick questions, one technical and one policy oriented:

techish: Is there a standardized reference model for agencies, either in some code format (perhaps compiled by NIST here), or by a third party?  How could one refer to the US DOJ, for example, in an internationally reliable framework?

wonkish: What sort of standards could one use to determine when a piece of legislation has become noteworthy?  It seems that there's a real opportunity to create an automated way to designate "hot" or perhaps relevant legislation, by maybe blog traffic, amount of comments, recent legislative progress, news coverage, or perhaps some combination of those.           If that's a weak distinction (noteworthy versus not noteworthy), then what would be a useful set of ways to classify legislation?  Perhaps procedural, party-line, novel, introduced for the 23rd time, well debated vs rushed, etc...   (see especially opencongress.org 's "hot legislation" for an example of this idea in action)

--
John Wonderlich

Program Director
The Sunlight Foundation
(202) 742-1520 ext. 234

[Tom Bruce]

On Jun 7, 7:16 pm, "John Wonderlich" <johnwonderl...@gmail.com> wrote:
> Two quick questions, one technical and one policy oriented:
>

> *techish*: Is there a standardized reference model for agencies, either in

> some code format (perhaps compiled by NIST

> here<http://www.itl.nist.gov/fipspubs/by-num.htm>),

> or by a third party? How could one refer to the US DOJ, for example, in an
> internationally reliable framework?

John:
I'm a little confused by what you mean by "standard reference model"
-- "reference model for what purpose?" is the question that comes to
mind.

That's not going to stop me from asserting that there isn't any that
I know of, though ;-). Part of the question is whether the model you
want models structure or not. The only reference I know of that works
that way is the US Government manual, and it incorporates no unique
identifier scheme that I can detect. So far as flat structures go,
the Federal Register and FDMS use some standardized reference
identifiers for docket numbers in rulemakings, but I don't think they
extend any further than those units that actually do N&C rulemaking.

I have got to think that OMB, GAO, and GSA must have some kind of
internal identifier scheme for a variety of purposes, probably
accounting.

So far as "internationally reliable" goes, again, the question is
purpose. Telex addresses or domain names would work for some
applications ;-). If it's taxonomy or any kind of transnational-
comparative thing you're interested in, things bog down in apple-and-
orange problems very quickly because of radically different approaches
to government organization.

So, I circle back to my original question: what do you want to do?


>
> *wonkish*: What sort of standards could one use to determine when a piece of

> legislation has become noteworthy? It seems that there's a real opportunity
> to create an automated way to designate "hot" or perhaps relevant
> legislation, by maybe blog traffic, amount of comments, recent legislative
> progress, news coverage, or perhaps some combination of those. If
> that's a weak distinction (noteworthy versus not noteworthy), then what
> would be a useful set of ways to classify legislation? Perhaps procedural,
> party-line, novel, introduced for the 23rd time, well debated vs rushed,
> etc... (see especially opencongress.org 's "hot legislation" for an
> example of this idea in action)

Deeply dependent on who's doing the noting, I'd say, particularly
given the increasing level of topical aggregation and stealth-
provision-insertion going on. Relevance measures might include
a) Viewer-declared interest in a particular statute or reg ("show me
everything affecting XX USC YYYY or NN CFR MM.PPP")
b) Viewer-set process threshold ("made it out of committee")
c) Viewer-set threshold for process delta ("changed status X times in
Y days")
d) Volume of comments submitted -- trouble here is setting an absolute
number because as you know they vary so widely in both number and
substantive usefulness
e) Ditto blog posts, news coverage, etc.
f) Viewer-declared interest in sponsor, agency, other governmental
unit, author, commentator, etc.
g) Whole thing to be trained and adjusted based on user's bookmarking
and clickstreaming behavior.

a) and b) are easy; f) is more or less easy depending on the quality
of the metadata attached to the thing itself; c) I don't know; d), e),
and g) probably depend initially on overinclusive text retrieval that
the user trains over time, not unlike spam filtering, or prying into
other applications ("show it to me if it's commented on by any of the
blogs in my aggregator that I read on a regular basis").

Multiple-factor approaches should allow the viewer to reweight factors
themselves.

I should think it would be very difficult to avoid echo-chamber
effects with this.