Semantic Web Rule Language (SWRL) Expressiveness Extensions

Tracking #: 1657-2869

Authors: 
Abba Lawan
Abdur Rakib
Natasha Alechina

Responsible editor: 
Thomas Lukasiewicz

Submission type: 
Survey Article
Abstract: 
SWRL is a direct extension of OWL designed to enable rule-based assertion of facts into OWL ontologies. In this survey, we explore the state-of-the-art of SWRL’s expressiveness extensions proposed over time. We set the stage by discussing the efficiency of the SWRL/OWL combination in conceptual domain knowledge modeling and further highlights the expressive limitations of the SWRL as a motivation. The majority of the paper discusses relevant language extensions, their added expressive powers to the original SWRL definition and a comparative analysis of their implementations where applicable. We further discuss the decidability requirements of the reviewed SWRL extensions, in relation to the basic DL-safety restriction of SWRL rules. We conclude the paper with our observations on the usability of SWRL rules and a table of summary categorizing the various extensions, which are mainly: the fuzzy and probabilistic extensions for managing uncertainties, the non-monotonic extensions and existentials, advanced mathematic extensions, and notable built-in extensions for added expressiveness to the classical SWRL definition.
Full PDF Version: 
Tags: 
Reviewed

Decision/Status: 
Reject

Solicited Reviews:
Click to Expand/Collapse
Review #1
Anonymous submitted on 14/Jun/2017
Suggestion:
Reject
Review Comment:

The paper reviews some SWRL extensions such as fuzzy, uncertainty and nono-monotonic ones.

Overall, I believe that a paper like this makes sense once a reasonable corpus of works have been published about a topic, such as the one here: SWRL extensions. However, right now we are speaking about a handfull of papers, which does not motivate at all this survey paper.

Below a list of minor issues, figured out here and there:

- authors name: I’m surprised to see them in lower case

- p.2: according to SWRL spec. in builtIn(r, x, …), x has to be bound to a dataLiteral.

- p.3: strictly speaking swrlx: makeOWLThing(..) isn’t ins the SWRL spec. and even the semantics is quite obscure…

- p.5: FOL,
a) which can express almost everything
b) though is undecided in terms of computational complexities such as space and time

Concerning a): what do you mean here?
Concerning b): the sentence is just scientifically ill formed. E.g., a decision problem may be undecidable or decidable e.g., in polynomial space/time

-p.6: use latex \sqcap, \sqcup for intersection and union of concepts

- p.6: sub-profiles -> profiles

- p.6, right column: â˘A ¸S -> ???

- p.15: Prposed -> Proposed

- p.16: ’Formula’: uses appropriate quoting ‘: `Formula’. Occurs several times

- References: check citations. Too many contain errors, are incorrect/incomplete or using non-conventional citation style, such as:
[1,2,6,7,21,24,33,35,39,42]

Review #2
Anonymous submitted on 12/Jul/2017
Suggestion:
Reject
Review Comment:

The paper surveys extensions of the Semantic Web Rule Language (SWRL).
It considers Fuzzy and Probabilistic Extensions, Non-monotonic Extensions, Rules
Ordering and Priority Extensions, Existential Extensions, Advanced Mathematical
Support Extensions and Built-in Extensions.

While the coverage is wide, the paper suffers from a lack of formality and
precision. For example, I would have expected a formal definition of the
semantics of basic SWRL and formal definitions of how the semantics of the
various extensions differ from it: instead, semantics are only described
informally and often very briefly.
The paper also contains errors, such as
Page 5
"Decidability: In the context of this work, decidability
refers to the ability of a ’Reasoner’ to
achieve inference — checking the consistency
of logical consequences of ontology axioms and
return true or otherwise, within a reasonable
amount of time"
decidability does not poses an upper limit on the amount of time, it only
imposes that it is finite
Similarly in
"FOL, which can express almost everything,
though is undecided in terms of computational
complexities such as space and time"

The paper does not clarify what is the current status of research
on SWRL extensions: most references date back to pre 2010. Given that
SWRL never became an W3C standard, is research on SWRL and its extensions still
active? Why a survey on such topic is important?
Moreover, OWL2 is treated only in passing, while it is the current OWL standard:
what is the relationship of OWL2 with SWRL? This is only hinted at in Figure 2.

The paper does not draw connections with neighboring fields where
similar issues have been discussed for years. Non-monotonic
extensions such as negation as failure have been amply discussed in logic
programming, where semantics for negation have achieved a mature status:
when discussing non-monotonic extensions, a comparison with such semantics
is necessary. When discussing existential quantifiers in heads,
Datalog +/- should
have been mentioned and compared with the presented extensions.
The problem of integrating numerical reasoning in rules has been amply studied
in the field of constraint logic programming.

The bibliography has issues:
reference [32] has an incorrect list of authors, an incorrect journal and an
incorrect doi:
instead of
J. M. A. Calero, A. M. Ortega, G. M. Perez, J. A. B.
Blaya, A. F. G. Skarmeta, M. Günther, T. Wiemann, S. Albrecht,
J. Hertzberg, J. Zhang, L. Zhang, S. Rockel, B. Neumann,
J. Lehmann, K. S. R. Dubba, A. G. Cohn, A. Saffiotti,
F. Pecora, M. Mansouri, Š. Koneˇcný, M. Günther,
S. Stock, L. S. Lopes, M. Oliveira, G. H. Lim, H. Kasaei,
V. Mokhtari, L. Hotz, W. Bohlken, S. Rudolph, A Nonmonotonic
Expressiveness Extension on the Semantic Web
Rule Language, Artificial Intelligence 11 (May) (2011) 1–36.
doi:10.1016/j.artint.2014.12.007.
I think it should be
Jose M. Alcaraz Calero, Andrés Muñoz Ortega, Gregorio Martínez Pérez, Juan A. Botía, Antonio Fernandez Gómez-Skarmeta:
A Non-monotonic Expressiveness Extension on the Semantic Web Rule Language. J. Web Eng. 11(2): 93-118 (2012)
The doi above points at
Martin Günther, Thomas Wiemann, Sven Albrecht, Joachim Hertzberg, Model-based furniture recognition for building semantic object maps, Artificial Intelligence, Volume 247, June 2017, Pages 336-351, ISSN 0004-3702, https://doi.org/10.1016/j.artint.2014.12.007.
(http://www.sciencedirect.com/science/article/pii/S000437021400157X)
which is completely unrelated
List of issues:
Page 2
"While the
DL-safety may result in an incomplete deduction of
knowledge in a given ontology, inferences from DLsafe
rules are always formally sound."
DL-safety does not result in incomplete deduction, it makes the language less
expressive so some things are not expressible.
Page 2
"atoms, which can contain a combination of
OWL constructs and axioms" how can axioms appear in rules?
Page 3
"only those concepts or
terms previously defined in the ontology can be used
in the SWRL rules" what does it mean for a concept to be defined in the
ontology? You can use class expressions as atoms in rules, and class
expressions define concepts
Page 3:
"SWRL axioms" undefined, you defined only SWRL rules
Page 5:
"The DL Expressiveness Algorithm (SHOIN(D)):" what does this mean?
What is the expressiveness algorithm? Is SHOIN(D) this algorithm?
Page 6:
The syntax and semantics of F-SWRL is given only briefly and informally:
for example, the paper does not explain how the truth values of atoms (which
are in [0,1]) are combined with the weights attached to atoms in rules.
Page 7: Vague-SWRL is not understandable from your description, for example
rule (2) does not adhere to the syntax you define
Page 9: f-NSWRL: the example "Person
p is rich and p is definitely not hungry but the health
status of p is not known" does not match rules (4) and (5)
In tables 1 and 2, what is the meaning of S(P)?
In equation (8), what is the use of Married(?m)? ?m does not appear
elsewhere in the rule
In equation (9), hasSpouse has now arity one, what is the meaning of
notExists(hasSpouse(?p))? According to your description, it should always
be false, as ?p is guaranteed to exist because of the atom Participant(?p)
Similarly for notExists(bookingS tatus(?b; "SinglesOnly")) in eq (10):
?b is guaranteed to exist
Page 14
"Do not construct
a rule, which has existentially quantified variables,
to form acyclic chain between its atoms or with
atoms from other rules": acyclic chain is undefined. Besides, below you say
"rules need to be tracked manually to ensure that their
executions will not lead to a cyclic chain of existential
quantification" so the chain must not be cyclic or acyclic?
Equation (15): there is no relationship between ?g and ?m
Rule (16) can be expressed in DL (OWL):
Workgroup\sqsubseteq \exists hasMember.Professor
Page 16: what is the difference between (i) and (iii)
Page 19: (i) and (ii) together mean that there must not be any chain,
whether cyclic or acyclic


Comments