HOME
Our Blog
Link to us
Bug Reports
The BlogWire
Local Keywords
Add a blog

Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
DC
Delaware
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming

State Parties
Elected Officials
Democracy for...
Labor Movement
Western Democrats

One suggestion, one bug report.

Let people put invisible tags in their posts to mark them as local.

<local state="NH" /> would work, but maybe it'd be safer to simply do something like <span id="nh" /> although, maybe <span state="nh" /> is even cleaner. Maybe you could check for any of the above.

--

Bug report:

If you check NH, you'll see a post near the top from "One Voice From NH". It is obviously from last year, though, but the date of the year is correctly placed.

--

I can help if you have a SQL or database question. Movable Type is a php thing, so that's not tough for me either. Tons of experience. ;) Check my blog for a link to my resume.

Josh-- Interesting idea. That said, I'm not sure I want to require people to use invalid HTML - even if it is for a good cause.

On the NH bug report - yeah, every once in a while a date goes askew. Seems to be less than 0.1% of the time.

Thanks for the offer of help. We're squared away for now, but I've got your resume on file now. :)

Hi, This post which is strictly local, ended up being listed as non-local Massachusetts.

Three Winners in MA State Senate Special Election Primary
Frederick Clarkson - Tue 8:08 p.m. - 5 clicks

Not being to be a downer, but what is the point of this local vs non-local business? Do I have to type "Ohio" in every post to get it listed properly? My blog has "Ohio" in it's title; why isn't this enough to be listed as local?

TickedOff...

Because that's the WHOLE point this website. There are many, many, many other aggregators and indexes out there.

People are having trouble finding local CONTENT about local issues.

We're trying to separate the I-Hate-Karl-Rove stuff from the Impeach-Bob-Taft stuff. There's precious little local content out there, but a real hunger for it.

My initial instinct was to ignore all the non-local posts completely, but because the system isn't perfect, we're running it all - just categorized.

Soon, when I'm done organizing this, then I'll be working on a Bayesian filter to automagically 'learn' over time what's local and non-local.

Remember, we're beta. And it's been less than three weeks. Hang in there.

<span id="NH" /> is 100% valid XHTML Strict 1.0

The only trick with ID tags is that they should be unique within a document.

<span class="NH" /> works great, unless the person had previously defined a CSS class named NH, in which case they could use "New Hampshire" or "____NH" with any random number of _'s.

The other suggestion wasn't valid XHTML or HTML, but it wouldn't have broken any browser anywhere :)

Maybe not the right place to post, but there are about 10 posts in the Ohio not local category that are listed as being from my site Ohio Watch. This problem is that they aren't from Ohio Watch they are from http://www.soapblox.net/myleftwing which is a national blog. Just a heads up.

Kari,

I guess I should ask what exactly it is you consider to be local, if you don't think calling for the impeachment of the governor of a state is local news. Are you planning on drilling down to the county level for all fifty states?

People are having trouble finding local content because blogs focusing on local activities don't exists at least in Ohio. I suggest that all those having trouble either read the newspaper or start their own local blogs.

Let me recap just one more time. The local/not-local thing is an AUTOMATED SYSTEM. I don't personally review all the thousands of posts a day.

If you use the name of the Governor in the headline or the RSS excerpt of your post, it'll wind up in the local feed. Sometimes, people use nicknames or abbreviations which the system doesn't know.

As I've said before, we've only been live a few weeks. We're working on a better system, that will allow the readers to flag posts as local or not-local.

Hang in there.

I've worked with Bayesian classifiers. My suggestion is to think of something else.

And I'd settle for aggregating local *blogs* vs local *content*, even if they do say they hate Rove every once and a while.

The present system certainly needs some help.

Well, now, hang on a minute. If you do indeed want to see all posts from all local blogs, then just click on the "combined" tab. That's what it's there for.

Most other people I've talked to want to find local content - so that's what that tab is for.

We're creating options here - not cutting them off.

p.s. Bayesian systems work great if they're trained right. It is, of course, possible to write a bad Bayesian classifier - which ones have you worked with, and how were they bad?

How terrible would it be to allow visitors to reclassify entries on the fly? That way if someone saw a post that was misclassified, they could click a button to get it switched. After a certain number of 'votes' from unique IPs, then the item would switch its classification.

That might even help you train your classification algorithm faster.

SCO -- that's precissely what we're working on. Great minds think alike!

Viz the classifier: Yes, good training data is key. If you use the "pages" I assume your archiving from each blog, make sure you have some way to manually classify/review the training data. What you don't want to do is to use the current categories of local vs. non-local since they have too many mistakes.

(Having some means for readers to reclassify as suggested above could be used for marking training data, but if noone chooses to do it for a given state, then you just have the current set-up. Perhaps you should explicitly reach out to local bloggers to see if they will help you do that.)

And this won't be a one time process, since you'll have to retrain since new terms will crop up as time passes which will be high probability indicators for the "local" category. For example, in the Nevada section, the current system mis-categorized a post about Richard Perkins, the speaker of the Nevada State Assembly (and about as local as you're going to get) because (I guess) Perkins isn't one of your "nevada local" keywords. But as time goes on, Perkins may not be important to Nevada politics anymore, but some new guy will be.

So one thing you might do is to make sure you get and update at least a list of all the state level officials and give those names extra weight in your scheme.

For the short-term, I would suggest making the "combined" button the default "view" of the state blogs. Otherwise, many truly local posts are mistakenly hidden. Those who want to risk the "local" button view could select that if they want .

Another computer geek's chiming in: read this thread and I think I see and agree with where you're going -- bayes plus user-supplied adjustments, right?

If so, an intermediate step might be to allow users to submit words that are local/nonlocal differentiators. A step easier than bayes, and the code framework could be done in a way that lets you just extend things to add bayesian filters as phase two.

Then again, the story I'm staring at is a nasty test case. A columnist that writes an even split of local/non-local editorials is caught in a political controversy for criticizing a local candidate for the inappropriate politicization of the race and for the candidate's utter lack of experience and their capitalizing on their celebrity in an unrelated national issue campaign.
The story's definitely local, but using either one's name to indicate local (what a user would do) will backfire when either name resurfaces on a national issue.

Anyway, you've made a great site. I just found it and will be hanging out here regularly, I imagine.

LeftyBlogs is the place to find out what's happening in the progressive blogosphere across the country, in every state.

© 2006 Mandate Media