W3C home > Mailing lists > Public > public-semweb-lifesci@w3.org > March 2009

Re: Is OWL useful at all for Quantitative Science?

From: Phillip Lord <phillip.lord@newcastle.ac.uk>
Date: Tue, 31 Mar 2009 11:47:17 +0100
To: "Matthias Samwald" <samwald@gmx.at>
Cc: "Oliver Ruebenacker" <curoli@gmail.com>, "public-semweb-lifesci" <public-semweb-lifesci@w3.org>
Message-ID: <81bprinf4a.fsf@newcastle.ac.uk>
"Matthias Samwald" <samwald@gmx.at> writes:
>>  I have tried to come up with a simple example. Feel free to come up
>> with a simpler one:
>>
>>  Express in correct OWL: Washington DC is further away from Boston
>> than New York City
>>
>>  Use case: I want to fly with my helicopter from Boston to either DC
>> or NYC, whichever is closer.
>
> Why should this be hard? If I take your example by word and I am free to come
> up with arbitrary OWL DL, we could simply use an n-ary design pattern to SAY
> it in OWL. E.g., create a class "is farther away than", with three properties
> "reference place", "nearer place", "place that is farther away" -- and create
> an instance accordingly. Problem solved.

I think I would say problem represented rather than solved. This is a
good thing, of course, and does represent a way in which you can do
quantitative science where OWL is part of the solution. 

Katy Wolstencroft did the same thing in this paper

doi:10.1093/bioinformatics/btl208 

In this case, the results of a quantitative analysis ("how much
similarity") were translated into statements in OWL, then we applied
reasoning over the top. 

At Newcastle, Keith Flanagan did a similar thing looking for genomic
rearrangements, which you can read about here:

http://homepages.cs.ncl.ac.uk/phillip.lord/download/publications/iswc2005_bayesian_poster.pdf

In this case, we were using OWL to help plug resources into a Bayesian
stats engine, and later interpret the results. 

Bottom line with all of these, is that can interact between OWL and
quantitative results. In each of these cases, the degree of interaction
between the numerical reasoning and logical reasoning was through a
fairly thin pipe. 

> But I guess what we really would want to do is to describe each city with
> geo-tags (latitude and longitude). Then we can use SPARQL to query for cities
> and calculate their distance from Boston.

Didn't know you could do that; is the mathematics integrated into SPARQL
or are you doing some kind of call out. The reason that I ask is that
ultimately, your question is "which is closer as the helicopter flies"
which you can only really get from a database, what with nofly zones and
the like.

Phil
Received on Tuesday, 31 March 2009 10:48:08 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:20:41 UTC