In message 
<[log in to unmask]
uk>, "REYNOLDS, Trevor" <[log in to unmask]> writes
>If I get this right, one of the advantages of linked data for me, could 
>be that I would no longer have to maintain my own Places Authority file 
>when we put our collections on-line.  Instead we could use someone 
>else's?

Good question.  I think that the answer depends on the context in which 
you expect to use the external place authority.

In general terms, the main advantage of using an external place 
authority file in a Linked Data context is that your statements about 
places can be "understood" more widely, because you are using the same 
place identifiers as other data providers.  Your own place authority 
identifiers are meaningless outside your own system.

Another potential advantage is that the external authority may contain 
additional information (such as co-ordinates), which means that you get 
"added value" from using it.

Lets say that you already have a working system which includes 
references to your own place authority file.  If you are thinking of the 
possibility of replacing all those references by references to, say, 
Geonames [1] identifiers or Ordnance Survey geopolitical entities [2], 
then you have to consider how that change might be achieved.

In an ideal world this would be an automatic process; in practice it 
would probably involve a considerable amount of manual intervention.  In 
particular, I don't think that it is wise to expect that you will be 
able to start from your own existing data, and routinely do an 
auto-lookup of the corresponding Linked Data authority entry, on the 
fly.  Instead, you should expect to have to record Linked Data 
identifiers in your own records, either instead of, or in addition to, 
your existing place data.  How automatically you can add those Linked 
Data identifiers is another question.

Another approach might be to add statements to your existing place 
authority file to say that your place X is "the same as" Geonames place 
Y. (Or you could keep these equivalence assertions separately.)  This 
approach involves less work, since you don't have to update individual 
records.  However, it does mean that you have to stick with your 
existing authority file!

>Now let's look at some practical issues.
>
>First London, within our own system we try to be clear on whether we 
>mean the City of London, Greater London or the old London County 
>Council Area.  What sort of information would one need in the Linked 
>Data world in order to ensure that one matched up with the right 
>London? (and indeed not London. Ontario, London, Mississippi etc.).

This example nicely illustrates the point that single place names are 
not, of themselves, sufficient to identify geographical or geopolitical 
areas unambiguously.  (This is equally true of personal names.)  In 
order to be sure that an external place authority entry is the right 
one, you will need to match on a number of properties.  These might 
include place type, names of containing places, co-ordinates, etc.  If 
you search Geonames for an exact match on "London", with the country 
code set to "GB", you get two hits:

<geonames style="MEDIUM">
<totalResultsCount>2</totalResultsCount>
<geoname>
<name>London</name>
<lat>51.5084152563931</lat>
<lng>-0.125532746315002</lng>
<geonameId>2643743</geonameId>
<countryCode>GB</countryCode>
<countryName>United Kingdom</countryName>
<fcl>P</fcl>
<fcode>PPLC</fcode>
</geoname>
<geoname>
<name>City of London</name>
<lat>51.5133363996235</lat>
<lng>-0.0890064239501953</lng>
<geonameId>2643744</geonameId>
<countryCode>GB</countryCode>
<countryName>United Kingdom</countryName>
<fcl>A</fcl>
<fcode>ADM2</fcode>
</geoname>
</geonames>

Without specifying the "GB" property, you get 55 hits.

Making such a match automatically requires (a) that you have recorded 
enough properties in your own data and (b) that the external authority 
and your own data record such properties in a compatible manner.  This 
is why, in practice, I would expect there to be an element of manual 
checking or selection involved.  In the above case, you need to 
understand Geonames' <fcl> and <fcode> coding in order to decide which 
"London" to use.

>For the second sort of object let us look at 79703652.121 an object in 
>our collection described as a Geological Specimen from St Paul's 
>Island. Now there is a St Paul's Island in Malta but there are also - 
>at least - St Paul Islands in (1) Nova Scotia, Canada (in the Gulf of 
>St Lawrence). (2) Alaska, USA (in Priboff Islands). (3) Amsterdam and 
>St Paul District, French Southern and Antarctic Lands.    Not knowing 
>which of these four places our particular rock comes from we currently 
>link it in our database to "St Paul's Island (unlocated place)" a 
>record which has a note describing the confusions.  How would you deal 
>with this sort or undefined data in the Linked Data world?

Well, you could do exactly the same thing in a Linked Data authority 
(which is no different in its expressive capabilities than your own 
authority file: it's just a database exposed/expressed in a particular 
way). Conversely you could accept that all you know about the collection 
place is that it was called St Paul's Island, and record just that fact.

Richard

[1] http://www.geonames.org/
[2] http://data.ordnancesurvey.co.uk/
-- 
Richard Light

****************************************************************
For mcg information visit the mcg website at
http://museumscomputergroup.org.uk/
To manage your subscription to this email list visit
http://museumscomputergroup.org.uk/email-list/
****************************************************************