<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Bounded Implication for Existential Rules (Extended Abstract)</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Cristina Civili</string-name>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Riccardo Rosati</string-name>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>DIAG, Sapienza Universita` di Roma</institution>
          ,
          <country country="IT">Italy</country>
        </aff>
        <aff id="aff1">
          <label>1</label>
          <institution>School of Informatics, University of Edinburgh</institution>
          ,
          <country country="UK">UK</country>
        </aff>
      </contrib-group>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>-</title>
      <p>
        The problem This paper deals with the property of boundedness in rule languages.
Boundedness is an important notion that formalizes the fact that a rule set can be
unfolded into a finite set 0 of acyclic (i.e., non-recursive) rules such that and 0
are equivalent on every database: it is therefore a crucial property for optimizing the
processing of rules. Such a property has been extensively studied, especially for the
Datalog rule language [
        <xref ref-type="bibr" rid="ref4 ref9">9,4</xref>
        ], and, recently, for Answer Set Programming [
        <xref ref-type="bibr" rid="ref16">16</xref>
        ].
      </p>
      <p>
        In Datalog, the (uniform) boundedness of a program P can be defined as the
existence of an integer k such that, for every database D, the number of iterated applications
(in a forward chaining manner) of P to D that are necessary to compute the minimal
model of P and D is bounded by k. This definition of boundedness is equivalent to
the existence of a finite, non-recursive program that is equivalent to P . Also, it is
wellknown that a Datalog query is bounded if and only if it is equivalent to a first-order
sentence [
        <xref ref-type="bibr" rid="ref1 ref12">1,12</xref>
        ].
      </p>
      <p>
        More recently, rule-based languages have been used in the context of
ontologybased data access [
        <xref ref-type="bibr" rid="ref15">15</xref>
        ]. In this framework, the main focus is on the problem of
answering conjunctive queries over an ontology expressed by a set of rules, and one of the
most studied properties is the first-order rewritability of conjunctive queries
(CQFOrewritability) over an ontology, which corresponds to the above mentioned first-order
expressibility in Datalog: an ontology O is CQFO-rewritable if every conjunctive query
q over the ontology can be equivalently rewritten into a first-order query q0, i.e., q0 is
such that, for every database instance D, the evaluation of q over O and D coincides
with the evaluation of q0 over D. Notably, in the case when the ontology is expressed
as a set P of Datalog rules, the CQFO-rewritability of P and the boundedness of P are
equivalent properties.
      </p>
      <p>
        Existential rules, which extend Datalog rules to the presence of existentially
quantified variables and multiple atoms in rule heads, have been proposed and studied in the
last years as a specification language for ontology-based data access [
        <xref ref-type="bibr" rid="ref14 ref2 ref5">5,2,14</xref>
        ]. An
existential rule (or simply rule) over a relational schema S is an expression of the form
8x8y( (x; y; a) ! 9z (x; z; b)), where (x; y; a) (the body of ) and (x; z; b) (the
head of ) are conjunctions of atoms over S. We call x the frontier variables (F ( )),
and z the existential head variables of (EH ( )), while a; b are the constants
occurring in (Const( )). Several recent studies have focused on the first-order rewritability
property for existential rules (e.g., [
        <xref ref-type="bibr" rid="ref2 ref7 ref8">7,2,8</xref>
        ]). On the other hand, the notion of
boundedness for existential rules has not been deeply investigated. To our knowledge, one
of the most relevant recent approaches to this problem is presented in [
        <xref ref-type="bibr" rid="ref2">2</xref>
        ], where the
notion of acyclic graph of rule dependencies (aGRD) is defined, which corresponds
to a form of boundedness for existential rules. Instead, we start our study from a
notion of boundedness for existential rules that generalizes the definition of boundedness
provided for Datalog to existential rules in a simpler way: we call such a notion strict
boundedness. More precisely, we say that a set of existential rules is strictly bounded
if it is logically equivalent to a finite and acyclic set of rules.
      </p>
      <p>However, it can be immediately verified that, for arbitrary sets of existential rules,
the above notion of boundedness is much stronger than the first-order rewritability of
conjunctive queries. That is, while strict boundedness of a rule set implies its
CQFOrewritability, there exist rule sets that are CQFO-rewritable but are not strictly bounded.
Notice that the same property holds for the above mentioned notion of aGRD.</p>
      <p>
        The main goal of this paper is to answer the following question: is it possible to
generalize the notion of boundedness for Datalog to existential rules, in such a way that
the correspondence with the notion of first-order rewritability of conjunctive queries
is preserved? Actually, from the forward chaining perspective, such a generalization
has been provided by the bounded derivation depth property of the chase of existential
rules [
        <xref ref-type="bibr" rid="ref10 ref5">5,10</xref>
        ]. However, we would like to characterize this property in terms of equivalent
representations of the set of rules, and see how the alternative notion of boundedness as
existence of a finite and non-recursive equivalent rule set has to be extended to capture
first-order rewritable rule sets.
      </p>
      <p>
        Our contribution Our approach to the study of boundedness for existential rules is
inspired by the work in query rewriting for existential rules [
        <xref ref-type="bibr" rid="ref11 ref13 ref2 ref6">2,13,6,11</xref>
        ]. In particular,
we extend the techniques presented in [
        <xref ref-type="bibr" rid="ref13 ref2">2,13</xref>
        ] to address the problem of computing an
unfolding of a set of existential rules, and the problem of defining an appropriate notion
of redundancy between rules.
      </p>
      <p>First, we define a notion of boundedness for existential rules that weakens strict
boundedness by giving up the acyclicity condition: we call such a notion weak
implication-boundedness. It is based on the idea of looking for a finite representation
of all the single-head rules (i.e., rules with one atom in the head) that are logically
implied by the initial rule set, and on a notion of equivalence between rule sets, called
R-equivalence, that is different from (and stronger than) the standard logical
equivalence. R-equivalence is based on a notion of redundancy between two rules.</p>
      <p>Given two rules : (x; y; a) ! 9z (x; z; b), and 0 : 0(x0; y0; c) !
9z0 (x0; z0; d), we say that is redundant with respect to 0 if there exists a
specialization of 0 ( 0) = s0 (where : F ( 0) ! F ( 0) [ Const( )) and a bijective
function : F ( s0) ! F ( ) such that the following first-order sentences are valid:
8x8y body( )(x; y; a) ! 9y0 (body( s0))(x; y0; e)
8x08z0 head( s0)(x0; z0; d) ! 9z (head( ))(x0; z; f )
where e = c [ a [ b and f = d [ a [ b.</p>
      <p>Given two sets of rules , 0, we say that 0 R-entails if, for each
nontautological rule 2 there exists a rule 0 2 0 such that is redundant w.r.t.</p>
      <p>0. Moreover, we say that and 0 are R-equivalent if both 0 R-entails and
R-entails 0.</p>
      <p>Let be a set of rules over a signature S. We define the SH-closure of as the set
?s = f j is a single-head rule over S and Const( ), and j= g.</p>
      <p>We say that a set of rules is weakly implication-bounded if ?s is R-equivalent
to a finite set of rules.</p>
      <p>It is immediate to verify that strict boundedness always implies weak
implicationboundedness, while the converse does not always hold. Furthermore, it turns out that
weak implication-boundedness is not equivalent to CQFO-rewritability. More precisely,
it is possible to show that weak implication-boundedness does not imply
CQFOrewritability for single-head ternary set of rules (i.e., rules over relations of arity 3).
Similarly, it can be shown that weak implication-boundedness does not imply
CQFOrewritability for binary set of rules (i.e., over relations of arity 2). On the other hand,
we are able to prove the correspondence between weak implication-boundedness and
the notion of AFO-rewritability, that is, first-order rewritability of all atomic queries
(i.e., conjunctive queries consisting of a single atom).</p>
      <p>To arrive at a notion of boundedness that corresponds to CQFO-rewritability, we
define a second notion of strong implication-boundedness for existential rules. Roughly
speaking, such a notion is obtained from weak implication-boundedness by
discarding the restriction to single-head rules in the deductive closure of the rule set, and by
considering projections of such a deductive closure.</p>
      <p>Let be a set of rules over a signature S. We call closure of the set ? =
f j is a rule over S and Const( ), and j= g. Moreover, let ; 0 be
two rules. We say that 0 is head-unifiable w.r.t. if there exists a homomorphism
: F ( ) ! F ( 0) [ Const( 0) and an isomorphism : EH ( ) ! EH ( 0) such that
head( ( ( ))) = head( 0). Then, we call projection of a rule set with respect to a
rule the set ( ) = f 0 2 j 0 is head-unifiable with g.</p>
      <p>We say that a set of rules over a schema S is strongly implication-bounded if, for
each rule over S, ( ?) is R-equivalent to a finite set of rules.</p>
      <p>
        The notion of weak implication-boundedness has the desired correspondence with
the CQFO-rewritability. Namely, every set of rules is strongly implication-bounded
if and only if is CQFO-rewritable. This in turn implies the correspondence between
strong boundedness and the bounded derivation depth property [
        <xref ref-type="bibr" rid="ref10">10</xref>
        ].
      </p>
      <p>
        Moreover, the equivalence between weak and strong implication-boundedness
actually extends to two broad classes of existential rules: single-head binary rules, that
is, single-head rules over relations of arity not greater than 2, and frontier-guarded [
        <xref ref-type="bibr" rid="ref2">2</xref>
        ]
rules.
      </p>
      <p>We believe that the equivalence between weak and strong implication-boundedness
is a very important property for a set of existential rules. In particular, the above
correspondence could be exploited in the optimization of query answering over ontologies
expressed by rule sets belonging to the above classes.</p>
      <p>
        Finally, it is possible to show that checking strong (or, equivalently, weak)
implication-boundedness is undecidable for single-head binary rules, and (using
results from [
        <xref ref-type="bibr" rid="ref3">3</xref>
        ]) decidable for frontier-guarded rules. These results complement the
wellknown undecidability of (strict) boundedness for Datalog [
        <xref ref-type="bibr" rid="ref9">9</xref>
        ].
      </p>
      <p>Acknowledgments This research is partially supported by the UK EPSRC Project
VADA (grant n. M025268) and the EU Project Optique (grant n. FP7-318338).</p>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          1.
          <string-name>
            <surname>Miklo</surname>
          </string-name>
          <article-title>´s Ajtai and Yuri Gurevich. Datalog vs first-order logic</article-title>
          .
          <source>J. Comput. Syst. Sci.</source>
          ,
          <volume>49</volume>
          (
          <issue>3</issue>
          ):
          <fpage>562</fpage>
          -
          <lpage>588</lpage>
          ,
          <year>1994</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          2. Jean-Franc¸ois Baget, Michel Lecle`re,
          <string-name>
            <surname>Marie-Laure Mugnier</surname>
            , and
            <given-names>Eric</given-names>
          </string-name>
          <string-name>
            <surname>Salvat</surname>
          </string-name>
          .
          <article-title>On rules with existential variables: Walking the decidability line</article-title>
          .
          <source>Artificial Intelligence</source>
          ,
          <volume>175</volume>
          (
          <fpage>9</fpage>
          -10):
          <fpage>1620</fpage>
          -
          <lpage>1654</lpage>
          ,
          <year>2011</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          3.
          <string-name>
            <given-names>Vince</given-names>
            <surname>Ba</surname>
          </string-name>
          <article-title>´ra´ny, Michael Benedikt, and Balder ten Cate. Rewriting guarded negation queries</article-title>
          .
          <source>In Mathematical Foundations of Computer Science 2013 - 38th International Symposium, MFCS 2013</source>
          , pages
          <fpage>98</fpage>
          -
          <lpage>110</lpage>
          ,
          <year>2013</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          4.
          <string-name>
            <given-names>Michael</given-names>
            <surname>Benedikt</surname>
          </string-name>
          , Balder ten Cate, Thomas Colcombet, and Michael Vanden Boom.
          <article-title>The complexity of boundedness for guarded logics</article-title>
          .
          <source>In 30th Annual ACM/IEEE Symposium on Logic in Computer Science, LICS</source>
          <year>2015</year>
          , Kyoto, Japan, July 6-
          <issue>10</issue>
          ,
          <year>2015</year>
          , pages
          <fpage>293</fpage>
          -
          <lpage>304</lpage>
          ,
          <year>2015</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          5. Andrea Cal`ı, Georg Gottlob, and
          <string-name>
            <given-names>Thomas</given-names>
            <surname>Lukasiewicz</surname>
          </string-name>
          .
          <article-title>A general datalog-based framework for tractable query answering over ontologies</article-title>
          .
          <source>Semantic Web J.</source>
          ,
          <volume>14</volume>
          :
          <fpage>57</fpage>
          -
          <lpage>83</lpage>
          ,
          <year>2012</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          6. Andrea Cal`ı, Georg Gottlob, and
          <string-name>
            <given-names>Andreas</given-names>
            <surname>Pieris</surname>
          </string-name>
          .
          <article-title>Advanced processing for ontological queries</article-title>
          .
          <source>Proc. of the VLDB Endowment</source>
          ,
          <volume>3</volume>
          (
          <issue>1</issue>
          ):
          <fpage>554</fpage>
          -
          <lpage>565</lpage>
          ,
          <year>2010</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          7. Andrea Cal`ı, Georg Gottlob, and
          <string-name>
            <given-names>Andreas</given-names>
            <surname>Pieris</surname>
          </string-name>
          .
          <article-title>Towards more expressive ontology languages: The query answering problem</article-title>
          .
          <source>Artificial Intelligence</source>
          ,
          <volume>193</volume>
          :
          <fpage>87</fpage>
          -
          <lpage>128</lpage>
          ,
          <year>2012</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          8.
          <string-name>
            <given-names>Cristina</given-names>
            <surname>Civili</surname>
          </string-name>
          and
          <string-name>
            <given-names>Riccardo</given-names>
            <surname>Rosati</surname>
          </string-name>
          .
          <article-title>A broad class of first-order rewritable tuple-generating dependencies</article-title>
          .
          <source>In Proc. of the 2nd Datalog 2.0 Workshop</source>
          ,
          <year>2012</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref9">
        <mixed-citation>
          9.
          <string-name>
            <given-names>Haim</given-names>
            <surname>Gaifman</surname>
          </string-name>
          , Harry G. Mairson, Yehoshua Sagiv, and
          <string-name>
            <surname>Moshe</surname>
            <given-names>Y.</given-names>
          </string-name>
          <string-name>
            <surname>Vardi</surname>
          </string-name>
          .
          <article-title>Undecidable optimization problems for database logic programs</article-title>
          .
          <source>J. of the ACM</source>
          ,
          <volume>40</volume>
          (
          <issue>3</issue>
          ):
          <fpage>683</fpage>
          -
          <lpage>713</lpage>
          ,
          <year>1993</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref10">
        <mixed-citation>
          10.
          <string-name>
            <surname>Georg</surname>
            <given-names>Gottlob</given-names>
          </string-name>
          , Stanislav Kikot, Roman Kontchakov, Vladimir V. Podolskii, Thomas Schwentick, and
          <string-name>
            <given-names>Michael</given-names>
            <surname>Zakharyaschev</surname>
          </string-name>
          .
          <article-title>The price of query rewriting in ontology-based data access</article-title>
          .
          <source>Artif</source>
          . Intell.,
          <volume>213</volume>
          :
          <fpage>42</fpage>
          -
          <lpage>59</lpage>
          ,
          <year>2014</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref11">
        <mixed-citation>
          11.
          <string-name>
            <surname>Georg</surname>
            <given-names>Gottlob</given-names>
          </string-name>
          , Giorgio Orsi, and
          <string-name>
            <given-names>Andreas</given-names>
            <surname>Pieris</surname>
          </string-name>
          .
          <article-title>Ontological queries: Rewriting and optimization</article-title>
          .
          <source>In Proc. of the 27th IEEE Int. Conf. on Data Engineering (ICDE</source>
          <year>2011</year>
          ), pages
          <fpage>2</fpage>
          -
          <lpage>13</lpage>
          ,
          <year>2011</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref12">
        <mixed-citation>
          12.
          <string-name>
            <surname>Phokion</surname>
            <given-names>G.</given-names>
          </string-name>
          <string-name>
            <surname>Kolaitis and Christos H. Papadimitriou</surname>
          </string-name>
          .
          <article-title>Some computational aspects of circumscription</article-title>
          .
          <source>J. ACM</source>
          ,
          <volume>37</volume>
          (
          <issue>1</issue>
          ):
          <fpage>1</fpage>
          -
          <lpage>14</lpage>
          ,
          <year>1990</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref13">
        <mixed-citation>
          13. Me´lanie Ko¨nig, Michel Lecle`re,
          <string-name>
            <surname>Marie-Laure Mugnier</surname>
          </string-name>
          , and Michae¨l Thomazo.
          <article-title>Sound, complete and minimal ucq-rewriting for existential rules</article-title>
          .
          <source>Semantic Web</source>
          ,
          <volume>6</volume>
          (
          <issue>5</issue>
          ):
          <fpage>451</fpage>
          -
          <lpage>475</lpage>
          ,
          <year>2015</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref14">
        <mixed-citation>
          14.
          <string-name>
            <surname>Markus</surname>
          </string-name>
          <article-title>Kro¨tzsch and Sebastian Rudolph</article-title>
          .
          <article-title>Extending decidable existential rules by joining acyclicity and guardedness</article-title>
          .
          <source>In Proc. of the 22nd Int. Joint Conf. on Artificial Intelligence (IJCAI</source>
          <year>2011</year>
          ), pages
          <fpage>963</fpage>
          -
          <lpage>968</lpage>
          ,
          <year>2011</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref15">
        <mixed-citation>
          15.
          <string-name>
            <surname>Antonella</surname>
            <given-names>Poggi</given-names>
          </string-name>
          , Domenico Lembo, Diego Calvanese, Giuseppe De Giacomo, Maurizio Lenzerini, and
          <string-name>
            <given-names>Riccardo</given-names>
            <surname>Rosati</surname>
          </string-name>
          .
          <article-title>Linking data to ontologies</article-title>
          .
          <source>J. on Data Semantics</source>
          , X:
          <fpage>133</fpage>
          -
          <lpage>173</lpage>
          ,
          <year>2008</year>
          .
        </mixed-citation>
      </ref>
      <ref id="ref16">
        <mixed-citation>
          16. Heng Zhang and Yan Zhang.
          <article-title>First-order expressibility and boundedness of disjunctive logic programs</article-title>
          .
          <source>In IJCAI 2013, Proceedings of the 23rd International Joint Conference on Artificial Intelligence</source>
          , Beijing, China,
          <source>August 3-9</source>
          ,
          <year>2013</year>
          ,
          <year>2013</year>
          .
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>