[Talk-us] GNIS POI populations

Minh Nguyen minh at nguyen.cincinnati.oh.us
Tue Jan 13 20:39:10 UTC 2015


On 2015-01-13 07:22, Serge Wroclawski wrote:
> Mihn,
>
> If we do any en-mass edit, there are a few things I think we want to consider:
>
> 1. Before anything else, we need to make sure it's community approved,
> source data and code examined and approved by the community.

Sure. The code isn't ready yet. My e-mail was more of an early-morning 
musing. :-)

In particular, I noticed that plenty of POIs have an additional 
`census:population` tag that takes the form "1234;2006", where 2006 is a 
year and 1234 is a number that usually doesn't match the value in 
`population`. I'd like to update `population` and replace 
`census:population` with `population:date`, which is far more common 
worldwide.

> 2. I think that in principle this is a good idea, but we'll also
> encounter some issues where census data shows that objects:
>
> a. Have moved
> b. Have been renamed
> c. Are present in one dataset but not the other
>
>
> We want to flag these for secondary processing (likely manual).
>
> 3. I think it'd be nice, if we're doing en mass mechanical edits, to
> connect objects to Wikidata

At the moment, boundary relations are mostly linked to Wikipedia, but 
unfortunately few POIs are part of those relations. (They'd have a role 
of "label", which I think is poorly named, but that's a topic for 
another day.) Since they come from TIGER, the boundary relations also 
generally have the official census name in `tiger:NAMELSAD`, which would 
make step 2 easier.

> While it makes sense to keep our own copy of various attributes (like
> town/city classifications) in OSM, if we connect the data with
> Wikidata, in the future we could pull the data out of Wikidata for
> what's classified as what.
>
> They're going to be more inclined to have this kind of data updated
> regularly than we are, so this seems like a good way of moving towards
> this model, even if today we don't have a way of retrieving the data
> from the two datasets and reconciling them.
>
>
> I'd be okay with #3 being left to a separate task, but I think it
> deserves some thought.
>
> - Serge
>
> On Tue, Jan 13, 2015 at 8:34 AM, Minh Nguyen
> <minh at nguyen.cincinnati.oh.us> wrote:
>> It looks like most of the place=city/town/village/hamlet POIs from GNIS are
>> tagged with 2000 Census populations in the population tag. These population
>> tags allow renderers to label places with font sizes corresponding to
>> population, which is a pretty common use case.
>>
>> I think we should consider a mechanical edit to update these tags to the
>> 2010 Census figures en masse. I've been updating individual places as I edit
>> them for other reasons, but this tag is most useful when its vintage is
>> consistent across the board.
>>
>> --
>> minh at nguyen.cincinnati.oh.us
>>
>>
>> _______________________________________________
>> Talk-us mailing list
>> Talk-us at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/talk-us
>
> _______________________________________________
> Talk-us mailing list
> Talk-us at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-us
>


-- 
minh at nguyen.cincinnati.oh.us




More information about the Talk-us mailing list