Allow non-exact dates

[alastair]

We have some data that has a non-exact date, e.g. for a composer, Born: circa 1860, Died: circa 1920.
This information should be able to be stored in the CE.

[ChristiaanScheermeijer]

@alastair do you have a suggestion about how this data could be stored?

[alastair]

@musicog might have some ideas about how these kinds of dates are normally stored in linked data

[musicog]

Sorry for the delayed response, I missed this originally.

Unfortunately I don't have any satisfying response here -- approximate / uncertain dates are still somewhat an open problem for Linked Data; and it's a big problem, particularly in early publishing.

In a previous project around publishing the Early English Books Online (EEBO) corpus as Linked Data, we ended up defining our own precise- and exact- properties, e.g. eeboo:precise-birth and eeboo:precise-death for birth and death dates; see https://github.com/oerc-elephat/preprocessed-elephant/blob/master/eeboo-ontology.ttl

What makes it more difficult is that different data sets have a huge variety of ways of expressing precise and approximate dates -- here's the (large) set of regular expressions we wrote to capture them from the EEBO corpus (p == precise, a == approximate, r == range)
https://github.com/oerc-elephat/preprocessed-elephant/blob/master/regex-author-year.csv
https://github.com/oerc-elephat/preprocessed-elephant/blob/master/regex-publication-date.csv

The other approach is to cheat -- e.g., dbpedia does things like specifying "1551-0-0" to mean "sometime in the year 1551", see the death date for Richard Sherry on this page: http://dbpedia.org/page/Richard_Sherry

But of course the exact year (or even decade) is sometimes in dispute or unknown, in which case this particular solution doesn't work.

Ideally, this is something that we should think about carefully, with input from a (music) domain expert.

[ChristiaanScheermeijer]

Duplicate #91

[alastair]

Not quite the same, #91 talks about "Born in 1790-04, but we don't know the day", and this one is "Born some time around 1790 (maybe some years before or after)", but I agree that they're closely related and we could track them in one ticket. If we do so, we should add notes from this ticket to the other one.