lack of expressivity
Galen
interaction
reasoning problems
complex ways
cases
human hand
biology
degree
tight complexity bounds
design
hierarchy
family
Foundational Model
statements
domain
formalism
model
technical problems
phalanx
procedure
limitations
order
DL components
components
expressive power
manner
false
vein
small part
DL axioms
anatomy
2010
role
computational properties
skeleton
aspects
Web
properties
acyclicity
finger
formal underpinning
tree model property
distal phalanx
human body
algorithm
complexity bounds
representation
organs
basic reasoning problems
constructs
consortium
first-order rules
practice
complexity results
decomposition
medical doctors
such objects
way
search
Ontology Language
hierarchical decomposition
number of parts
possible interactions
modelers
bone
additional conditions
conditions
https://doi.org/10.1007/978-3-642-14197-3_4
extension
problem
representation formalism
consequences
hand
combination
description graph
results
usage
lack
description
internal structure
semantic underpinnings
tree-like manner
DL ontologies
en
primitives
satisfiability problem
conditional statements
restriction
decidability
underpinnings
elements
Web Ontology Language
main goal
number
observations
science
structure
severe consequences
intended structure
decision procedure
structured objects
The Web Ontology Language (OWL) is a well-known language for ontology modeling in the Semantic Web [9]. The World Wide Web Consortium (W3C) is currently working on a revision of OWL—called OWL 2 [2]—whose main goal is to address some of the limitations of OWL. The formal underpinnings of OWL and OWL 2 are provided by description logics (DLs)[1]–knowledge representation formalisms with well-understood formal properties.DLs are often used to describe structured objects—objects whose parts are interconnected in complex ways. Such objects abound in molecular biology and the clinical sciences, and clinical ontologies such as GALEN, the Foundational Model of Anatomy (FMA), and the National Cancer Institute (NCI) Thesaurus describe numerous structured objects. For example, FMA models the human hand as consisting of the fingers, the palm, various bones, blood vessels, and so on, all of which are highly interconnected.Modeling structured objects poses numerous problems to DLs and the OWL family of languages. The design of DLs has been driven by the desire to provide practically useful knowledge modeling primitives while ensuring decidability of the core reasoning problems. To achieve the latter goal, the modeling constructs available in DLs are usually carefully crafted so that the resulting language exhibits a variant of the tree-model property [10]: each satisfiable DL ontology always has at least one model whose elements are connected in a tree-like manner. This property can be used to derive a decision procedure; however, it also prevents one from accurately describing (usually non-tree-like) structured objects since, whenever a model exists, at least one model does not reflect the intended structure. This technical problem has severe consequences in practice [6]. In search of the “correct” way of describing structured objects, modelers often create overly complex descriptions; however, since the required expressive power is actually missing, such descriptions do not entail the consequences that would follow if the descriptions accurately captured the intended structure.In order to address this lack of expressivity, we extended DLs with description graphs, which can be understood as schema-level descriptions of structured objects. To allow for the representation of conditional statements about structured objects, we also incorporated first-order rules [3] into our extension. In this way we obtain a powerful and versatile knowledge representation formalism. It allows us, for example, to describe the structure of the hand using description graphs, statements such as “if a bone in the hand is fractured, then the hand is fractured as well” using rules, and nonstructural aspects of the domain such as “a medical doctor is a person with an MD degree” using DLs.To study the computational properties of our formalism, we base the DL component on the \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$\mathcal{SHOIQ}^+$\end{document} description logic, as this DL provides the semantic underpinning of OWL 2. The resulting formalism is quite expressive, and it is unsurprising that it is undecidable. We investigate restrictions under which the formalism becomes decidable. In particular, we have observed that structured objects can often be described by a possibly large, yet bounded number of parts. For example, a human body consists of organs all of which can be decomposed into smaller parts; however, further decomposition will eventually lead to parts that one does not want or know how to describe any further. In this vein, FMA describes the skeleton of the hand, but it does not describe the internal structure of the distal phalanges of the fingers. The number of parts needed to describe the hand is therefore determined by the granularity of the hierarchical decomposition of the hand. This decomposition naturally defines an acyclic hierarchy of description graphs. For example, the fingers can be described by description graphs that are subordinate to that of the hand; however, the description graph for the hand is not naturally subordinate to the description graphs for the fingers. We used this observation to define an acyclicity restriction on description graphs. Acyclicity bounds the number of parts that one needs to reason with, which, provided that there are no DL axioms, can be used to obtain a decision procedure for the basic reasoning problems.If description graphs are used in combination with DL axioms, the acyclicity condition alone does not ensure decidability due to possible interactions between DL axioms, graphs, and rules [5]. To obtain decidability, we limit this interaction by imposing an additional condition on the usage of roles: the roles (i.e., the binary predicates) that can be used in DL axioms must be separated from the roles that can be used in rules. We developed a hypertableau-based [7] reasoning algorithm that decides the satisfiability problem for our formalism, together with tight complexity bounds.All proofs and additional decidability and complexity results for the case when DL axioms are expressed in \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$\mathcal{SHOIQ}^+$\end{document} can be found in [8].
reasoning algorithm
latter goal
OWL 2
objects
modeling primitives
MD degree
desire
Combining Description Logics, Description Graphs, and Rules
knowledge representation formalism
clinical science
such descriptions
10-12
owls
part
granularity
clinical ontologies
revision
molecular biology
rules
further decomposition
palm
bounds
blood vessels
goal
Semantic Web
language
chapter
World Wide Web Consortium
power
chapters
persons
ontology
example
2010-01-01
axioms
description logics
thesaurus
acyclicity condition
doctors
numerous problems
variants
modeling constructs
complex description
expressivity
https://scigraph.springernature.com/explorer/license/
National Cancer Institute Thesaurus
vessels
proof
2022-05-20T07:48
graph
FMA
formal properties
logic
body
Dickson
Lukose
dimensions_id
pub.1031554793
978-3-642-14197-3
Conceptual Structures: From Information to Intelligence
978-3-642-14196-6
doi
10.1007/978-3-642-14197-3_4
Springer Nature
Croitoru
Madalina
Springer Nature - SN SciGraph project
Sébastien
Ferré
Motik
Boris
Information and Computing Sciences
Computing Laboratory, University of Oxford, UK
Computing Laboratory, University of Oxford, UK
Computation Theory and Mathematics