=Paper=
{{Paper
|id=Vol-1879/paper23
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-1879/paper23.pdf
|volume=Vol-1879
}}
==None==
<pdf width="1500px">https://ceur-ws.org/Vol-1879/paper23.pdf</pdf>
<pre>
       Rewriting Queries with Negated Atoms

                Enrique Matos Alfonso and Giorgos Stamou

                National Technical University of Athens (NTUA)
                         gardero@image.ntua.gr



Abstract. The current paper have been accepted at the International Joint Con-
ference on Rules and Reasoning (RuleML+RR 2017). We focus on Query rewrit-
ing, a popular approach for ontology based data access and in general for first
order rewritable knowledge bases. The algorithms defined in the field are based
on conjunctive queries with no use of negation over the atoms that are part of
them. Also, the constraints present in the knowledge base are ignored in the pro-
cess of rewriting a query and they are only used to check the consistency of the
data.
In this paper, we study the problem of answering queries that allow negated
atoms. We developed a novel method to use classical rewriting techniques for
answering conjunctive queries with negated atoms. For a given conjunctive query
with some negated atoms, we propose an algorithm that finds a set of conjunc-
tive queries with no negated atom that contain all the answers of the initial query
with respect to the rules. The algorithm uses resolution with respect to the clauses
corresponding to the query and the constraints of the system in order to produce
rewritings of the initial query without negated atoms. Our approach uses a classi-
cal rewriting algorithm as a black box and the constraints in the system to find the
set of conjunctive queries without negated atoms that is equivalent to the original
query containing negated atoms.
A system (C OMPLETO) was implemented with the proposed method and com-
pared to another system (R EBSIR) that is able to rewrite negated concepts. In the
experimental evaluation C OMPLETO performed better than R EBSIR for most of
the datasets in the benchmark and it portrayed a more scalable performance i.e.
describing a faster relative performance with respect to R EBSIR’s performance
with the increase of the size of assertions in the dataset.

</pre>