[pedantic-web] Over-enthusiatic owl:sameAs linkage of Dailymed and LinkedCT data

14 views
Skip to first unread message

Hogan, Aidan

unread,
Apr 26, 2010, 10:24:41 AM4/26/10
to christi...@fu-berlin.de, pedant...@googlegroups.com
Hi Chris,

Please forward as appropriate ;).

I have some concern about the amount of owl:sameAs linkage between the
drug datasets ([1,2,3]) given in [1]. The concern is based on results
such as [4] for (offline/debug) SWSE results consolidated purely through
asserted owl:sameAs statements and the transitive/symmetric closure
thereof -- admittedly [4] is not much use for debugging the
consolidation, but gives an idea of the resulting mess :).

It seems that there is heavy owl:sameAs linkage specified by [1] to [2],
which is of course not necessarily a bad thing, but as far as I can
tell, it's gotten a little out of control.

For example, [5] gives a selection of sameas statements that link
various Dailymed resources to a specific LinkedCT drug [6]. I can make
an educated guess that groupings of equivalent drugs are being created
according to a shared value for dailymed:activeIngredient (in this case
Sodium Chloride), or possibly the owl:sameAs relation is only created if
*all* active ingredients match.

I'm not a drugs expert (aside from having read a few Hunter S. Thompson
books), but this does not seem like a strong enough case for saying that
two drugs are the same. This is especially the case when you consider
e.g., [7], where the list of active-ingredients is incomplete.

Aside from that, there are some strange sameAs statements not following
the active ingredient criteria mentioned before... E.g., between [5] and
[8].

Again, when you take all such sameAs links between all such drugs, and
stick transitive/symmetric closure in there, things quickly get out of
hand. E.g., [5] sameAs [8] sameAs [9]...

Along those lines, I would suggest weakening the owl:sameAs relation, or
strengthening the criteria for matching (about which, I admittedly can
only make educated guesses).

Cheers,
Aidan

[1] http://www4.wiwiss.fu-berlin.de/dailymed/
[2] http://data.linkedct.org/
[3] http://dbpedia.org/
[4]
http://deri-srvgal21.nuigalway.ie/swse/detail?focus=http%3A%2F%2Fdata.li
nkedct.org%2Fresource%2Fintervention%2F10009
[5]
http://www4.wiwiss.fu-berlin.de/dailymed/snorql/?describe=http%3A%2F%2Fd
ata.linkedct.org%2Fresource%2Fintervention%2F61884
[6] http://data.linkedct.org/page/intervention/61884
[7]
http://www4.wiwiss.fu-berlin.de/dailymed/snorql/?describe=http%3A%2F%2Fw
ww4.wiwiss.fu-berlin.de%2Fdailymed%2Fresource%2Fdrugs%2F1131
[8]
http://www4.wiwiss.fu-berlin.de/dailymed/snorql/?describe=http%3A%2F%2Fd
ata.linkedct.org%2Fresource%2Fintervention%2F61884
[9]
http://www4.wiwiss.fu-berlin.de/dailymed/snorql/?describe=http%3A%2F%2Fw
ww4.wiwiss.fu-berlin.de%2Fsider%2Fresource%2Fdrugs%2F3823








--
Subscription settings: http://groups.google.com/group/pedantic-web/subscribe?hl=en
Reply all
Reply to author
Forward
0 new messages