TY - GEN
T1 - What Causes Pneumonia? The Case for a Standard Semantics for ``may'' in OWL
AU - Rector, A
AU - Stevens, R
AU - Drummond, N
A2 - Dolbear, C
A2 - Ruttenberg, A
A2 - Sattler, U
A2 - Dolbear, C
A2 - Ruttenberg, A
A2 - Sattler, U
N1 - Long Paper in Volume 432
PY - 2008
Y1 - 2008
N2 - One of users' most frequent questions about OWL is ``how do I say `may'?'', as in ``Bacteria may cause pneumonia.'' In many fields, particularly biomedicine, a high proportion of knowledge that users wish to capture in ontologies is of this form. The issue is closely related to dealing with fillers other existential restrictions, most obviously in partonomies: ``All cars have engines'', therefore ``Engines may be parts of cars'', but not ``All engines are part of cars.'' There is no standard mechanism in the OWL language specification, API, reasoners, or query languages for dealing with these issues. In the absence of a standard, users have found various work arounds - some partial solutions, some clearly incorrect. In this paper we examine the intuitions to be captured and four such work arounds. We then sketch an approach to extending OWL and associated query languages that is a reasonable approximation to the intuitions, provides standard semantics so that there are standard answers to queries, and falls within the capability of existing reasoners, although it would require extensions to APIs and tools.
AB - One of users' most frequent questions about OWL is ``how do I say `may'?'', as in ``Bacteria may cause pneumonia.'' In many fields, particularly biomedicine, a high proportion of knowledge that users wish to capture in ontologies is of this form. The issue is closely related to dealing with fillers other existential restrictions, most obviously in partonomies: ``All cars have engines'', therefore ``Engines may be parts of cars'', but not ``All engines are part of cars.'' There is no standard mechanism in the OWL language specification, API, reasoners, or query languages for dealing with these issues. In the absence of a standard, users have found various work arounds - some partial solutions, some clearly incorrect. In this paper we examine the intuitions to be captured and four such work arounds. We then sketch an approach to extending OWL and associated query languages that is a reasonable approximation to the intuitions, provides standard semantics so that there are standard answers to queries, and falls within the capability of existing reasoners, although it would require extensions to APIs and tools.
M3 - Conference contribution
VL - 432
T3 - CEUR Workshop Proceedings
BT - Proceedings of the 5th OWLED Workshop on OWL: Experiences and Directions, collocated with the 7th International Semantic Web Conference (ISWC'08)
PB - RWTH Aachen University
ER -
