[Freebase-discuss] Dated types

[Raymond Yee]

I enjoyed Paul's discussion of A != A and his example of Poland.  In 
trying to clean up the US county data right now, I face exactly this 
problem.  Since I'm not certain about how to go forward, I'll sketch out 
a proposal below to get some feedback.

  Consider

http://www.freebase.com/view/en/skagway-hoonah-angoon_census_area

A bit of background:  I quote from 
http://www.census.gov/geo/www/tiger/ctychng.html :

    Hoonah-Angoon Census Area (02-105): Created from the remainder of
    the former Skagway-Hoonah-Angoon Census Area (02-232) when the new
    Skagway Municipality (02-230) was created effective June 20, 2007;
    population: 2,574.

I renamed this topic from Skagway-Hoonah-Angoon Census Area to 
Hoonah–Angoon Census Area to reflect the fact that the description and 
wikipedia link of 
/en/skagway-hoonah-angoon_census_area/en/skagway-hoonah-angoon_census_area 
correspond more to the current Hoonah-Angoon (and less to the historic 
Skagway-Hoonah-Angoon) -- see 
http://en.wikipedia.org/wiki/Hoonah%E2%80%93Angoon_Census_Area,_Alaska

However, my change doesn't clean up the problems.  (First, what to do 
about the id /en/skagway-hoonah-angoon_census_area and second, what 
about  Skagway-Hoonah-Angoon?)

  My proposal is

1) split this topic into two topics: one for Hoonah-Angoon and another 
for Skagway-Hoonah-Angoon

2) keep the /en/skagway-hoonah-angoon_census_area for the latter and 
name the former /en/hoonah-angoon_census_area

3) make sure to mark /en/skagway-hoonah-angoon_census_area as a dated 
location with a dissolution date.

4) write the appropriate FIPS 6-4 (now taken over as INCITS 31:200x -- 
http://www.census.gov/geo/www/ansi/ansi.html) for both counties. Maybe 
we should retain the FIPS 6-4 code for the defunct Skagway-Hoonah-Angoon 
Census Area

5) Consider some type of name change for the name authority.  
Technically, FIPS 6-4 (along with some other FIPS standards)  is no 
longer an active standard and has been replaced by corresponding ANSI 
standards: http://www.census.gov/geo/www/ansi/ansi.html    I sense that 
most people who know about FIPS standards will continue to think about 
FIPS codes -- but what do about keys whose names should ch change?  Add 
new keys?

Does my proposal sound right? If so, do I flag this topic for splitting 
and then wait for the split to happen. Once it happens, is it the 
staff's job to then set up the proper id for the two topics?

-Raymond


On 5/6/10 7:08 AM, Paul Houle wrote:
> Ed Laurent wrote:
>    
>> There have been many(!) discussions about time mediation of just about
>> everything over the past few years so there is a general need for
>> linking time to topics.
>>
>> But, some things to consider ...
>> 1) quoting jg on a recent post "we are trying not to be a temporal database".
>> 2) lack of data can indicate a lack of data as well as "current" status
>> 3) There is often a desire to maintain current data as distinct
>> properties for queries, resulting in denormalization.
>>
>> I'm personally all for maintaining historical data in Freebase but am
>> still puzzled as to how to efficiently do it outside of dated
>> properties in personal bases, which make it difficult to combine
>> historical data among types.
>>
>> -Ed
>>
>>      
> There are quite a few strategies for this. You can qualify statements,
> you can qualify predicates, you can qualify subjects, etc.
>
> Historical modelling gets tough quick. Consider the vernacular entity
> "Poland"; the recorded history of this region starts around 966 AD, but
> the actual "shape" of Poland has varied drastically since then. At times
> it has covered a much wider land area than it has now, but there have
> also been times that it hasn't existed at all (as a state.) How do you
> represent something like this? One of the principles that I'm chewing on is
>
> A != A
>
> That is, no term is atomic. There's a vernacular conception of "Poland"
> but then there are specific "Polish" states that have existed at
> different times, such as Polish-Lithuanian Commonwealth, the Jangiellon
> dynasty, etc. No database is going to be the 'last word' on reality, but
> just a starting place for deeper and deeper explorations of the truth.
>
> If you're going to tackle the issue of time-dependent times, how about
> time-dependent names? There's a prison a few miles from my house and it
> seems like they change the name on the signs that point to it every year
> or so. Fierce battles have been fought, in both the real world and
> wikipedia, about a city that Germans call Danzig and Poles call Gdańsk.
> Wikipedians ultimately agreed on time-dependent names for the city:
>
> http://en.wikipedia.org/wiki/Talk:Gdansk/Vote
>    

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freebase.com/pipermail/freebase-discuss/attachments/20100507/741f1157/attachment.htm