Introduction: Catherine Dolbear

Dear all,
I'd like to introduce myself to the list, along with my colleague Jenny
Green, who will also be attending some of the conference calls, and the
use cases we have at Ordnance Survey.

Ordnance Survey is the national mapping agency of Britain, and has a
large spatial database, whose semantics we are both expressing and
enhancing with ontologies [1]. To convert all our data directly to RDF
would result in about 50 billion triples (more if we explicitly encode
spatial relationships), so we are investigating how to map from a
semantic to a relational model. The two main drivers of our research are
1) content customisation (describing a customer's knowledge and
terminology in an ontology, and linking it to Ordnance Survey data, thus
allowing them to query using their own words, not ours, for geographical
objects in our data) and 2) semantic data integration (merging two
ontologies, one of which is linked to our relational database, and one
to a customer's data). Both of these require the mapping of an ontology
to a relational database (we are using Oracle 10g)

Some of the problems we are currently having are:
1) The semantic gap. 
By this I mean that the attribution in our data is not as rich as the
domain-level ontology, that is, the way users of the data view the
world. For example, the term "Island" is not represented explicitly in
our data, but can be derived from those rows in the Topographic Area
table that have a Theme column value of "Land" and are spatially inside
other rows with a Theme column of "Water". This means that the
simplistic option of mapping one class to a table, and a property to a
column is a bit useless: we need far more sophisticated mappings. (I
believe this is generally going to be the case for most applications -
for any sort of repurposing and customisation of data you need to move
beyond the original semantics in the database.)  At the moment we have
encoded these mappings as OWL, but SWRL rules, or some other standard
format, may be better. Deciding what these mappings are is also pretty
difficult, and a good user interface, visualisation and navigation is
also really important.

2) Dealing with spatial queries.
Although this may be a minority use case, many people will need to
consider calculations or dealing with data that's not just strings, for
example images etc. At the moment we are mapping the RCC-8 spatial
relationships which appear either in the mapping ontology or the SPARQL
query directly to the Oracle spatial operators, as they have a
one-to-one correspondence. I'm not sure how this could be used in a more
general case however, or be non-database specific though?

Some more discussion of our work can be found in [2]. 

We're looking forward to working with you, and hopefully addressing some
of these issues!

Catherine

Links
[1] www.ordnancesurvey.co.uk/ontology 
[2] http://www.w3.org/2007/03/RdfRDB/papers/dolbear.pdf 


Dr Cathy Dolbear
Senior Research Scientist
Ordnance Survey Research 
www.ordnancesurvey.co.uk/research 

.


This email is only intended for the person to whom it is addressed and may contain confidential information. If you have received this email in error, please notify the sender and delete this email which must not be copied, distributed or disclosed to any other person.

Unless stated otherwise, the contents of this email are personal to the writer and do not represent the official view of Ordnance Survey. Nor can any contract be formed on Ordnance Survey's behalf via email. We reserve the right to monitor emails and attachments without prior notice.

Thank you for your cooperation.

Ordnance Survey
Romsey Road
Southampton SO16 4GU
Tel: 08456 050505
http://www.ordnancesurvey.co.uk

Received on Thursday, 10 April 2008 15:20:42 UTC