[DL] Description Logic Online Seminar November Edition

[Bartosz Bednarczyk]

Dear members of the DL community,
We are pleased to invite you to the upcoming Description Logics Seminar,
which will be held next Friday, November 15, 2024, at 2:00 PM via Zoom.
This session features two exceptional speakers: Maurice Funk from Leipzig
University and Anton Gnatenko from the Free University of Bozen-Bolzano.
All further details about the talks will be available here:
https://dl.kr.org/seminar
For convenience, abstracts for their talks are attached below
Looking forward to seeing you there!
Best regards,
Bartosz Bednarczyk (on behalf of the DL Seminar Organizing Team)


Talk by Maurice Funk (Leipzig University)
Learning Conjunctive Queries under Description Logic Ontologies
Learning conjunctive queries (CQs) from labeled data examples requires
deciding NP-hard problems, even for very restricted classes of CQs. One way
to circumvent this and enable learning in polynomial time, is to give the
learner access to more information through membership queries. This notion
of learning is formalized by Angluin's exact learning model. This talk is
concerned with the exact learning of CQs in the setting where queries are
answered under a description logic (DL) ontology. I will give an overview
of several results on the learnability of queries under ontologies
formulated in DLs from the EL and DL-Lite families, as well as the
implications of these results on the learnability of DL concepts in the
same setting. Additionally, I will point out the main open questions in
this area.

Talk by Anton Gnatenko (Free University of Bozen-Bolzano)
First-Order Rewritability of Ontology-Mediated Queries in (Temporal)
Description Logics
An ontology-mediated query consists of a TBox T and a database query Q.
Sometimes, all the reasoning required to answer (T, Q) can be "packed" into
an existential positive first-order formula \phi, called the first-order
rewriting, so that answering (T, Q) is equivalent to just evaluating \phi,
over any ABox. Evaluating first-order formulae is fast and can be done by
standard database engines. Therefore, if \phi exists, it is preferable to
compute it once and reuse it when you need to answer (T, Q), rather than
invoking a slower DL reasoner every time. The question is how to check
whether it exists and, if so, how to obtain it. I will survey how this
problem is (un)resolved for some description logics, and how it becomes
more complex in the context of logics with temporal operators.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.zih.tu-dresden.de/pipermail/dl/attachments/20241105/10af9121/attachment.htm>