[Data-modeling] Modeling uncertainty

[Scott Blomquist]

Very interesting suggestion. Is there any reason we should make the range of
uncertainty able to be +x/-y instead of +/-x? I.e. I can imagine a case
where someone's date of birth is known to be around a certain time, but it
could have been as much as 3 years earlier, but no more than 6 months later.
I realize that serves to make things more complex, but I think it's a more
accurate reflection of many kinds of date uncertainty. I think you're right
about using a number and unit of time, though. If we accept dates, I fear
that some data might get mistakenly entered thinking that this replaces the
estimated start or end dates in the Event type itself.

On Mon, Feb 16, 2009 at 3:24 PM, Kirrily Robert <kirrily at metaweb.com> wrote:

> I wonder whether, for events at least, we could have a co-type for
> uncertainly dated events that specify a range of uncertainty.  Jeff would
> kick me for this (if he weren't off having a baby) but call it something
> like "Uncertainly timed event" and have two properties, "Degree of
> uncertainty of start date", and "Degree of uncertainty of end date".  These
> expect a CVT which is an integer and a unit of time, eg. 3 days, 6 months,
> 1000 years.  For "Spring 1985" assuming it were the northern hemisphere, you
> could just put in a date of 1 May 1985 and allow 6 weeks' uncertainty, or
> thereabouts.
> This would leave you able to put an estimated date in the ordinary date
> fields on event, allowing it to appear neatly in timelines and whatnot, but
> also provide the information about the degree of uncertainty.
>
> K.
>
> On 16/02/2009, at 3:09 PM, Scott Blomquist wrote:
>
> I just encountered a scenario today that would benefit from the same
> solution as your date example. I found some events whose times I've only
> been able to pin down so far to "Spring 1985", and I don't think I have any
> good way to represent that in an event today.
>
> On Mon, Feb 16, 2009 at 10:28 AM, Tom Morris <tfmorris at gmail.com> wrote:
>
>> Is anyone doing work on modeling uncertainty?  I'm specifically
>> interested in dates and locations.
>>
>> Location - If I'm told that something is "near" or "in the vicinity
>> of" a location, currently my choices are to either not record the fact
>> or to guess at a way to reduce the precision in a way that's still
>> accurate.  I could say that something which is "near Boston" is "in
>> Massachusetts," but a) that might not be true and b) that's not the
>> information that I have.
>>
>> Dates - The simple case is "circa," but it would also be useful to
>> deal with both open and closed ranges (e.g. before 1945, after 1999,
>> or September 2008-December 2008).  Currently the only type of range
>> that can be encoded is ones which can be made by truncating precision
>> (ie 2009 == 1 Jan 2009 - 31 Dec 2009).
>>
>> Tom
>> _______________________________________________
>> Data-modeling mailing list
>> Data-modeling at freebase.com
>> http://lists.freebase.com/mailman/listinfo/data-modeling
>>
>
>
>
> --
> http://scott.blomqui.st
> _______________________________________________
> Data-modeling mailing list
> Data-modeling at freebase.com
> http://lists.freebase.com/mailman/listinfo/data-modeling
>
>
> --
> Kirrily Robert
> Freebase Community Director
> kirrily at metaweb.com
>
>
>
>
> _______________________________________________
> Data-modeling mailing list
> Data-modeling at freebase.com
> http://lists.freebase.com/mailman/listinfo/data-modeling
>
>


-- 
http://scott.blomqui.st
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freebase.com/pipermail/data-modeling/attachments/20090216/50136b29/attachment.htm