SlideShare uma empresa Scribd logo
1 de 32
Data Complexity in EL family of Description
                  Logics

                       Adila A. Krisnadhi

     (Joint work with Dr. Carsten Lutz (TU Dresden/Univ. Bremen)


                               2007
Introduction

      Increased interest in lightweight description logics (DLs)
          Admits tractable reasoning in large scale ontologies.
          EL family: tractable ontology language; sufficient expressive power
          for modeling life-science ontologies
          DL-Lite family: tailored towards applications with massive amount of
          instance data
      Most relevant reasoning services:
          instance checking: decision problem asking whether an individual is
          an instance of a concept/class w.r.t. background knowledge base
          (KB)
          conjunctive query answering: search problem related to a generalized
          form of instance checking
      Relevant complexity measure: data complexity (measured w.r.t. the
      size of instance data only), rather than combined complexity
      (measured w.r.t. the whole input: instance data, KB schema, query)
Objective

      Aim: investigate data complexity for the EL family
      We show data (in)tractability for wide range extensions of EL in the
      following. Data tractable = polynomial data complexity.
      Data-intractable (coNP-hard, already for instance checking):
            EL∀r .⊥ , EL∀r .C , ELC D
                            +
            EL∃¬r .C , EL∃r .C , EL∃r ∪s.C
            EL≤kr , EL≥kr for some fixed integer k ≥ 0
            ELkf : EL + k-functional roles (i.e., at most k successors), k > 1
      (Baader,et.al.,2005): the above DLs are at least PSpace-hard
      (most are ExpTime-hard) regarding combined complexity.
      Data-tractable (for conjunctive query answering):
            ELI f : EL + inverse role + (1)-functional role
            (Baader,et.al.,2005): Regarding combined complexity, ELI is
            PSpace-hard and ELf is ExpTime-complete
            (Hustadt,et.al.,2005): ELI f is data-tractable for instance checking
            Data-tractability results for DL-Lite family (Calvanese, et.al., 2006)
EL syntax & semantics


      Syntax: based on set of concept names NC and role names NR
      Set of (EL)-concepts (i.e., class in OWL):
          A is a concept for every A ∈ NC
            is a concept and if C , D are concepts, then C    D, ∃r .C are
          concepts too for r ∈ NR
      Semantics: Based on an interpretation I = (∆I , ·I ) where the
      domain ∆I is a set of individuals and ·I maps each concept to a set
      of individuals in ∆I and each role to a binary relation on ∆I .
          AI ⊆ ∆I for each A ∈ NC
          r I ⊆ ∆I × ∆I for each r ∈ NR
             I
               = ∆I
          (C D)I = C I ∩ D I
          (∃r .C )I = {x | ∃y : (x, y ) ∈ r I and y ∈ C I }
EL knowledge base (syntax)

   Consists of
       a TBox T : a set of concept definitions and/or concept inclusions
                                               .
            concept definition: statement otf A = C , A concept name, C
            concept
            concept inclusion: statement otf C D, C , D are concepts
       Two kinds of TBoxes:
            acyclic TBox: only allow concept definitions, LHS of all definitions
            are unique, has no cyclic definition
            general TBox: may also allow concept inclusion
       an ABox A: a set of concept assertions and role assertions
            concept assertion: expression otf A(a), A concept name, a individual
            name
            role assertion: expression otf r (a, b), r role name, a, b individual
            names
       All individual names are taken from a set NI .
EL KB (semantics)



     In an interpretation I, each individual name a ∈ NI is mapped to a
     domain individual aI ∈ ∆I
     I = (∆I , ·I ) satisfies the statements/expressions:
             .
         A = C iff AI = C I
         C D iff C I ⊆ D I
         A(a) iff aI ∈ AI
         r (a, b) iff (aI , b I ) ∈ r I
     I satisfies (i.e., a model of) a TBox T and an ABox A iff it satisfies
     all statements and expressions in them
Reasoning problems

      Satisfiability: “given a concept C and a KB (T , A), is there a model
      I of the KB that satisfies C , i.e., C I = ∅?”
      Subsumption: “given concepts C , D and a KB, is C I ⊆ D I for
      every model I of the KB?”
      KB Consistency: “given a KB, does the KB have a model?”
      Instance checking: “given an individual name a, a concept C and a
      KB K, is aI ∈ C I for every model I of the KB?” (written
      K |= C (a))
      These reasoning tasks can (usually) be reduced to each other.
      Satisfiability and KB consistency in EL is trivial; subsumption and
      instance checking in EL is tractable, even with a larger language
      (EL++ )
      Instance checking (and conjunctive query answering) emphasize
      reasoning over individuals and ABoxes, hence data complexity.
Conjunctive query answering & entailment
      Conjunctive query: a set q of atoms otf C (v ) and r (u, v ); u, v
      variables
           Given an interpretation I and a mapping π that maps the set of
           variables in q to ∆I , I satisfies C (v ) (resp. r (u, v )) w.r.t. π iff
           π(v ) ∈ C I (resp. (π(u), π(v )) ∈ r I )
           I satisfies q w.r.t. π (written I |=π q) iff I satisfies all atoms in q
           w.r.t. π
           I satisfies q (written I |= q) iff I |=π for some π
           Given knowledge base K: K |= q iff for every model of K, we have
           I |= q.
      Conjunctive query entailment: given K and q, decide whether K |= q
      Conjunctive query answering: given K and q, find all tuples of
      individual names in K such that when variables in q are properly
      substituted with the individual names, we have that K |= q
      Instance checking is a special case of conjunctive query entailment
      (i.e., with a single atom)
      Conjunctive query answering is the search problem corresponding to
      conjunctive query entailment.
Data intractibility results: overview

       All data-intractabilty results (coNP-hardness) are for instance
       checking w.r.t. acyclic TBoxes.
       The matching upper bound (in coNP) for most results is obtained
       from (Grimm,et.al.,2007) for SHIQ
       If the TBox is empty, most cases become data tractable: since
       ABoxes contain only concept names, either no query contains
       complex concept constructor (thus trivially reduced to query
       answering in EL which is tractable), or can be shown directly (the
       ELC D case).
       Exception: ELkf , k ≥ 2, instance checking is coNP-complete already
       when the TBox is empty.
       Some cases distinguish whether unique name assumption (UNA) is
       adopted.
           UNA: for two individual names a, b, a = b iff aI = b I
2+2-SAT


     Data intractability: coNP-hardness results by reduction from known
     NP-hard problem;
     (Schaerf, 1993): data complexity of instance checking for EL¬A is
     coNP-hard, by reduction from the NP-complete problem 2+2-SAT
     2+2-SAT: decide whether a given 2+2-formula is satisfiable
     2+2-formula: a finite conjunction of 2+2-clause
     2+2-clause: a propositional logic formula otf p1 ∨ p2 ∨ ¬n1 ∨ ¬n2
     where each disjunct is a propositional letter or a truth constant 1, 0.
     Reduction from 2+2-SAT to (negation of) instance checking for
     several extensions of EL will show coNP-hardness of instance
     checking for the corresponding DL.
General method for reduction



      Let ϕ = c0 ∧ · · · ∧ cn−1 be 2+2-SAT formula containing m
      propositional letters q0 , . . . , qm−1 .
      Let ci = pi,1 ∨ pi,2 ∨ ¬ni,1 ∨ ¬ni,2 for all i < n.
      Define a TBox T , an ABox A and a query concept C such that
          Size of A depends polynomially on size of ϕ
          Size of T and C are constant
      Show that A, T |= C (f ) (for some individual name f ) iff ϕ is
      satisfiable
EL¬A (Schaerf, 1993)
      EL¬A = EL + atomic negation (negation only on concept names).
           Semantics: (¬A)I = ∆I  AI
      Start from ϕ as defined earlier, define the following T , A and C :
                      .
              T := {A = ¬A}
              A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )}∪
                         {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )}
                   i<n

              C := ∃c.(∃p1 .A           ∃p2 .A      ∃n1 .A      ∃n2 .A)
      where f , 1, 0, q0 , . . . , qm−1 , c0 , . . . , cn−1 are individual names and c,
      p1 , p2 , n1 and n2 are role names. Note: size of A is polynomial in
      the size of ϕ
      A, T |= C (f ) iff ϕ is satisfiable. (See picture)
           Idea: C expresses that ϕ is not satisfied, i.e., there’s a clause in
           which the two positive literals and the two negative literals are false.
           (LR): Given a model I of A, T such that f I ∈ C I , show that a
                                                           /
           truth assignment that satisfies ϕ exists.
           (RL): Given a truth assignment satisfying ϕ, construct a model of
           A, T that does not satisfy C (f ).
EL∀r .⊥


          Essential technique from the previous proof: A and ¬A (in the
          TBox) partition the domain ∆I , i.e., every element in ∆I is either
          in AI or in (¬A)I .
          For EL∀r .⊥ (extension with ∀r .⊥), the partitioning concepts are
          ∃r . and ∀r .⊥
               Semantics: (∀r .bot)I = {x | ¬∃y : (x, y ) ∈ r I }
          Reduction proceeds like in EL¬A using the same ABox A and query
          concept C but the following TBox:
                                       .          .
                                T = {A = ∃r . , A = ∀r .⊥}

          Again, A, T |= C (f ) iff ϕ is satisfiable.
EL(≤kr ) , fixed k ≥ 1
       EL( ≤ kr ): extension with (≤ kr )
       No partitioning concepts; use covering concepts: ∃r .      and (≤ kr )
                              I                          I
            Semantics: (≤ kr ) = {x | #{y | (x, y ) ∈ r } ≤ k} where #S
            denotes cardinality of the set S.
       Reduction: ABox A and query concept C remains the same. TBox:
                                   .          .
                            T = {A = ∃r . , A = (≤ kr )}
       A, T |= C (f ) iff ϕ is satisfiable
            (LR): From a model I of A, T that doesn’t satisfy C (f ), define a
            truth assignment s.t. t(qi ) = 1 implies qiI ∈ AI and t(qi ) = 0
                            I
            implies qiI ∈ A . Such a truth assignment exist (due to covering),
            but needs not be unique since the covering concepts need not be
            disjoint. However, such a truth assignment always satisfies ϕ.
            (RL): Given a truth assignment t, define a model I of A, T that
            does not satisfy C (f ). Basically, the model resembles the ABox with
            additional individual d, and interprets A, A and r as follows:
                                                              I
                AI = {1} ∪ {qi | i < m and t(qi ) = 1}       A = {∆I  AI }
                r I = {(e, d) | e ∈ AI }
EL∀r .C , EL∃¬r .C



       EL∀r .C : extension with ∀r .C where C is a concept
           Semantics: (∀r .C )I = {x | ∀y : (x, y ) ∈ r I → y ∈ C I }.
       Reduction for EL∀r .C uses the covering concepts: ∃r .        and ∀r .X


       EL∃¬r .C : extension with ∃¬r .C where C is a concept
           Semantics: (∃¬r .C )I = {x | ∃y : (x, y ) ∈ r I and y ∈ C I }.
                                                     /
       Reduction for EL∃¬r .C uses the covering concepts: ∃r .           and ∃¬r .
+
ELC   D
          , EL∃r         .C


      ELC      D
                   : extension with C            D; (C        D)I = C I ∪ D I
      Reduction: use the following TBox, ABox and query concept:
                  .                  .     .
          T := {V = X            Y,A = X,A = Y}
          A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )} ∪ {V (qi ) | i < m}∪
                     {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )}
               i<n

          C := ∃c.(∃p1 .A           ∃p2 .A      ∃n1 .A      ∃n2 .A)
              +
      EL∃r .C : extension with ∃r + .C ;
      (∃r + .C )I = {x | ∃y : (x, y ) ∈ (r I )+ ∧ y ∈ C I } where the ’+’
      indicates the transitive closure of the corresponding role name.
      Reduction uses the same ABox and query concept as above and
                     .          .           .
      TBox T = {V = ∃r + .C , A = ∃r .C , A = ∃r .∃r + .C }
      Similar reduction can also be done for EL∃r ∪s.C .
EL(≥kr ) without UNA, k ≥ 2

      No two concepts a priori cover the domain. Hence, add more
      structures in the ABox.
      E.g. k = 3, reduction uses the same query concept C as before with
                         .           .
      the TBox T = {A = ∃r 4 . , A = (≥ 3r )} and ABox:

      A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )}
           ∪         {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )}
               i<n

           ∪         {r (qi , b1 ), r (qi , b2 , r (qi , b3 ), r (b1 , b2 ), r (b2 , b3 ), r (b1 , b3 )}
               i<m

      See picture
      Either two of b1 , b2 , b3 identify the same domain element or they do
      not. Hence, A and A defined in T provide the covering
EL(≥kr ) with UNA, k ≥ 2

      E.g. k = 3, reduction uses the same query concept C as before with
                          .         .              .
      the TBox T = {V = ∃r .B, A = ∃r .(A B), A = (≥ 3r )} and ABox:

            A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )}
                 ∪         {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )}
                     i<n

                 ∪         {r (qi , b1 ), r (qi , b2 , V (qi ), A(b1 ), A(b2 )}
                     i<m

      See picture
      In all models I of the KB, there is a d s.t. (qiI , d) ∈ r I for some
      i < m. Either d = bj (qi satisfies ∃r .(A B), or not (qi satisfies
      (≥ 3 r )).
      This reduction does not works without the UNA.
ELkf , k ≥ 2, without UNA

      This DL is EL extended with (global) k-functionality
            a role r is globally k-functional iff every element in the domain has
            at most k r -successor (i.e., satisfies the GCI      (≤ k r ))
      E.g., k = 2, reduction can be done without TBox and the following
      query concept and ABox:

      A := {r (1, e), A(e), B(e), r (0, e0 ), r (e0 , e1 ), r (e1 , e2 )}
            ∪ {c(f , c0 ), . . . , c(f , cn−1 )}
            ∪         {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )}
                i<n

            ∪         {r (qi , b1 ), r (qi , b2 ), r (qi , b3 ), r (b1 , b2 ), A(b1 ), A(b2 ), B(b3 )}
                i<m

      C := ∃c.(∃p1 .∃r 3 .              ∃p2 .∃r 3 .       ∃n1 .∃r .(A        B)     ∃n2 .∃r .(A   B))
ELkf , k ≥ 2, with UNA
      With UNA and without TBoxes, instance checking (and conjunctive query
      answering) is data-tractable.
          Consider the input ABox as complete description of an
          interpretation, check all possible matches of the query. (Taking into
          account possible inconsistency in the ABox).
      E.g., k = 3, reduction for instance checking with acyclic TBoxes:
              .
      T := {V = ∃r .B}
      A := {r (1, d1 ), r (1, d2 ), r (1, d3 ), s(1, d1 ), B(d1 )}
            ∪ {r (0, e1 ), r (0, e2 ), r (0, e3 ), s (0, e2 ), s (0, e3 ), B(d2 ), B(d3 )}
            ∪ {c(f , c0 ), . . . , c(f , cn−1 )}
            ∪         {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )}
                i<n

            ∪         {V (qi ), r (qi , bi,1 ), r (qi , bi,2 ), r (qi , bi,3 ), s(qi , bi,1 ), s (qi , bi,2 ), s (qi , bi,3 )}
                i<m

      C := ∃c.(∃p1 .∃s .B               ∃p2 .∃s .B        ∃n1 .∃s.B        ∃n2 .∃sB)

      qi satisfies either ∃s.B or ∃s .B
Lower bounds result summary



    EL extension                w.r.t. acyclic TBoxes      w.r.t. general TBoxes
    EL¬A                          coNP-complete               coNP-complete
    ELC D                         coNP-complete               coNP-complete
    EL∀r .⊥ , EL∀r .C             coNP-complete               coNP-complete
    EL(≤kr ) , k ≥ 0              coNP-complete               coNP-complete
    ELkf w/o UNA, k ≥ 2    coNP-complete (even w/o TBox)      coNP-complete
    ELkf with UNA, k ≥ 2   coNP-complete (in P w/o TBox)      coNP-complete
    EL(≥kr ) , k ≥ 2              coNP-complete               coNP-complete
    EL∃¬r .C                        coNP-hard                   coNP-hard
    EL∃r ∪s.C                       coNP-hard                   coNP-hard
         +
    EL∃r .C                         coNP-hard                   coNP-hard
ELI f : data-tractability overview

       ELI f : EL extended with inverse roles and functional roles
           inverse role: r − ; (r − )I = {(y , x) | (x, y ) ∈ r I }
           (globally) functional role; r is (globally) functional if it satisfies the
           GCI        (≤ 1 r )
           in ELI f , both a role r and its inverse can be declared functional.
           wlog. no inverse role in ABox and query concept.
       In this section, we consider general TBoxes.
       (Hustadt,et.al.,2005) Data-tractability results for Horn-SHIQ
       implies that instance checking for ELI f w.r.t. general TBoxes is
       data-tractable. Its direct proof is in my master’s thesis.
       We show the data-tractability result of conjunctive query answering
       for ELI f w.r.t. general TBoxes by giving a decision procedure for
       conjunctive query entailment running in polytime in the size of input
       ABox.
Assumptions
     We assume TBoxes are in normal form, i.e., GCIs are otf

        A     B,    A1    A2    B,    A     ∃r .B,    ∃r .A    B          (≤ 1 r )


            Every ELI f TBox T can be converted into normal form T in
            polytime by introducing fresh concept name.
            For every ABox A and conjunctive query q not using any of the
            concept names that occur in T but not in T , we have A, T |= q
     In all atoms C (v ) in a conjunctive query q, C is a concept name
     (i.e., no complex concept occurs in q)
            Can easily be achieved if C is not a concept name: replace C (v ) in q
                                  .
            with A(v ) and add A = C to the TBox where A is a concept name.
     Conjunctive queries are connected, i.e., for all variables u, v
     appearing in q, there are atoms r (u0 , u1 ), . . . , r (un−1 , un ) in q s.t.
     u = u0 and v = un
            Entailment of non-connected queries can be reduced to entailment of
            connected queries: if q is a non connected query, then A, T |= q iff
            A, T |= q for all connected components q of q (Glimm,et.al.,2007)
Algorithm


   Given an input TBox T , ABox A and conjunctive query q
       Convert T into normal form (polytime in |T |)
       If UNA is made, check consistency of A w.r.t. T (polytime in |A|
       due to (Hustadt,et.al.2005)). If inconsistent, answer “yes”.
       Otherwise, A must be admissible, and thus continue.
       If UNA is not made, convert A to make it admissible w.r.t. T by
       identifying individuals (polytime in |A|)
       Construct initial canonical structure I for T and A (polytime in
       |A|)
       Check matches of q against the above structure (polytime in |A|).
Admissible ABox w.r.t. TBox


      An ABox A is admissible w.r.t. a TBox T iff
          the UNA is made and A is consistent w.r.t. T ; or
          the UNA is not made and (        (≤ 1 r )) ∈ T implies that there are
          no individual names a, b, c occurring in A with r (a, b), r (a, c) ∈ A
          and b = c.
      Checking whether A is admissible w.r.t. T can be done in polytime
      in |A|:
          if the UNA is made, checking consistency of A w.r.t. T is equivalent
          to a number of instance checking w.r.t. T which is bounded in |A|
          if the UNA is not made, simply identify those individual names which
          are r -successors of some individual for all globally functional role r ;
          this is doable in polytime in |A|
Canonical structure
      Canonical model for A and T is the limit of the sequence of
      interpretations I0 , I1 , . . . defined as follows.
      Non-standard representation of interpretations is used: the function
      ·I maps every element d ∈ ∆I to a set of concept names d I ,
      instead of every concept name A to a set of elements AI .
      All interpretations satisfy
      ∆Ii ⊆ { a, p | a ∈ Ind(A) and p ∈ ex ∗ (T )} where
      ex ∗ (T ): the set of all paths (sequence of existentials) for T with ε
      the empty path
      ex(T ) is the set of all existentials (concepts ∃r .A occurring in the
      RHS of a GCI in T ) for T
      Let Γ be a finite set of concept names. If A ∈ Γ and A ∃r .B ∈ T ,
      then Γ has an ∃r .B-obligation O, where O contains
          B
          those concept names B in T such that there exists A ∈ Γ with
          ∃r − .A   B ∈T
          whenever r is globally functional: those concept names B such that
          ∃A ∈ Γ with A      ∃r .B ∈ T .
Canonical structure (2)
      Start from I0 defined as:
          ∆I0 = { a, ε | a ∈ Ind(A)}
          r I0 = {( a, ε , b, ) | r (a, b) ∈ A}
           a, ε I0 = {A ∈ NC | A, T |= A(a)}
          aI0 = a, ε
      Construct Ii+1 from Ii :
          If exists, select a, p ∈ ∆Ii and an α = ∃r .A ∈ ex(T ) s.t. a, p has
          α-obligation O, and
               r is not globally functional and a, pα ∈ ∆Ii ; or
                                                      /
               there is no b, p ∈ ∆Ii with ( a, p , b, p ) ∈ r Ii .
          Then do the following to get Ii+1 :
               add a, pα to ∆Ii
               if r is a role name, add ( a, p , a, pα ) to r Ii
               if r = s − , add ( a, pα , a, p ) to s Ii
               set a, pα Ii to subT (O), the closure of O under subsuming concept
               names w.r.t. T .
          Assumption: a, p is selected s.t. |p| is minimal (thus all obligations
          are eventually satisfied); set ex(T ) is well-ordered and the selected α
          is minimal for the node a, p , hence constructed canonical model is
          unique.
Important lemmas


     The canonical model I for T and A is a model of T and A.
     Let I be a canonical model for T and A, and J be a model for T
     and A. Then there is a homomorphism h from I to J s.t.:
         for all individual names a, h(aI ) = aJ ;
         for all concept names A and all d ∈ ∆I , d ∈ AI implies h(d) ∈ AJ ;
         for all (possibly inverse) roles r and d, e ∈ ∆I , (d, e) ∈ r I implies
         (h(d), h(e)) ∈ r J
     Let I be a canonical model for A and T , and q a conjunctive query.
     Then A, T |= q iff I |= q.
     The above lemmas show that we can decide query entailment by
     looking at only the canonical model, but the problem is that the
     canonical model is infinite.
Important lemmas (2)
      Idea: if we can show that the canonical model I satisfies q iff it
      satisfies q for some match π that maps all variables to elements
      reachable by traveling only a bounded number of role edges from
      some ABox individual, then we’re done.
      Let I be a canonical model for A and T . The initial canonical
      model I for A and T is obtained:
          ∆I = { a, p | |p| ≤ 2m + k}
          AI = AI ∩ ∆I
          r I = r I ∩ (∆I × ∆I
          aI = aI
      where m is the size of T and k is the size of q.
      Lemma: I |= q iff I |= q.
      Polytime (in |A|) construction of initial canonical model:
          I0 can be constructed in polytime in the size of A,
          obligations can computed in polytime because subsumption in ELI f
          is decidable and the required checks are independent of the size of A
          the number of elements of initial canonical model is polynomial in
          the size of A.
Summary

   EL extension                    w.r.t. acyclic TBoxes           w.r.t. general TBoxes
   EL¬A                             coNP-complete                     coNP-complete
   ELC D                            coNP-complete                     coNP-complete
   EL∀r .⊥ , EL∀r .C                coNP-complete                     coNP-complete
   EL(≤kr ) , k ≥ 0                 coNP-complete                     coNP-complete
   ELkf w/o UNA, k ≥ 2       coNP-complete (even w/o TBox)            coNP-complete
   ELkf with UNA, k ≥ 2      coNP-complete (in P w/o TBox)            coNP-complete
   EL(≥kr ) , k ≥ 2                 coNP-complete                     coNP-complete
   EL∃¬r .C                           coNP-hard                         coNP-hard
   EL∃r ∪s.C                          coNP-hard                         coNP-hard
        +
   EL∃r .C                            coNP-hard                         coNP-hard
   ELI f                                 in P                           P-complete

      For all considered extension, data-tractability can be shown iff the logic is convex
      regarding instances, i.e., A, T |= C (a) with C = D0 · · · Dn−1 implies
      A, T |= Di (a) for some i < n. (Can it be generalized?)
      Subtle differences such as the UNA or local vs. global functionality can have an
      impact on data-tractability.
Results published in ...



       A. Krisnadhi. Data Complexity of Instance Checking in the EL
       Family of Description Logics. Master’s thesis, Technische Universit¨t
                                                                          a
       Dresden, March 2007
       A. Krisnadhi & C. Lutz. Data Complexity in the EL family of DLs.
       In Proc.of the 20th Int. Workshop on Description Logics 2007
       (DL2007), p.88–99. 2007.
       A. Krisnadhi & C. Lutz. Data Complexity in the EL family of
       Description Logics. In Proc. of the 14th Int. Conf. on Logic for
       Programming, AI, and Reasoning (LPAR2007), vol. 4790 of LNAI,
       p. 333 – 347. 2007.
Referenced works

      F. Baader, S. Brandt, and C. Lutz. “Pushing the EL envelope”. In
      Proc. of the 19th Int. Joint Conf. on AI (IJCAI-05), pages 364–369.
      Morgan Kaufmann, 2005. (And its accompanying technical report).
      D. Calvanese, G. D. Giacomo, D. Lembo, M. Lenzerini, and R.
      Rosati. “Data complexity of query answering in description logics”.
      In Proc. of the 10th Int. Conf. on KR (KR06). AAAI Press, 2006.
      U. Hustadt, B. Motik, and U. Sattler. “Data complexity of
      reasoning in very expressive description logics”. In Proc. of the 19th
      Int. Joint Conf. on AI (IJCAI05), pages 466–471. Professional Book
      Center, 2005.
      B. Glimm, I. Horrocks, C. Lutz and U. Sattler. “Conjunctive Query
      Answering for the Description Logic SHIQ”. In Proc. of the 20th
      Int. Joint Conf. on AI (IJCAI-07). AAAI Press, 2007.
      A. Schaerf. “On the complexity of the instance checking problem in
      concept languages with existential quantification”. Journal of
      Intelligent Information Systems, 2:265–278, 1993.

Mais conteúdo relacionado

Mais procurados

Theory of Relational Calculus and its Formalization
Theory of Relational Calculus and its FormalizationTheory of Relational Calculus and its Formalization
Theory of Relational Calculus and its FormalizationYoshihiro Mizoguchi
 
Regular Languages
Regular LanguagesRegular Languages
Regular Languagesparmeet834
 
Theory of Computation "Chapter 1, introduction"
Theory of Computation "Chapter 1, introduction"Theory of Computation "Chapter 1, introduction"
Theory of Computation "Chapter 1, introduction"Ra'Fat Al-Msie'deen
 
Jarrar.lecture notes.aai.2011s.ch7.p logic
Jarrar.lecture notes.aai.2011s.ch7.p logicJarrar.lecture notes.aai.2011s.ch7.p logic
Jarrar.lecture notes.aai.2011s.ch7.p logicPalGov
 
Logistic Regression(SGD)
Logistic Regression(SGD)Logistic Regression(SGD)
Logistic Regression(SGD)Prentice Xu
 
Automata
AutomataAutomata
AutomataGaditek
 
CMSC 56 | Lecture 9: Functions Representations
CMSC 56 | Lecture 9: Functions RepresentationsCMSC 56 | Lecture 9: Functions Representations
CMSC 56 | Lecture 9: Functions Representationsallyn joy calcaben
 
3.1 intro toautomatatheory h1
3.1 intro toautomatatheory  h13.1 intro toautomatatheory  h1
3.1 intro toautomatatheory h1Rajendran
 
CommunicationComplexity1_jieren
CommunicationComplexity1_jierenCommunicationComplexity1_jieren
CommunicationComplexity1_jierenjie ren
 
Probabilistic information retrieval models & systems
Probabilistic information retrieval models & systemsProbabilistic information retrieval models & systems
Probabilistic information retrieval models & systemsSelman Bozkır
 
Regular language and Regular expression
Regular language and Regular expressionRegular language and Regular expression
Regular language and Regular expressionAnimesh Chaturvedi
 
Exchanging OWL 2 QL Knowledge Bases
Exchanging OWL 2 QL Knowledge BasesExchanging OWL 2 QL Knowledge Bases
Exchanging OWL 2 QL Knowledge Basesnet2-project
 

Mais procurados (18)

Theory of Relational Calculus and its Formalization
Theory of Relational Calculus and its FormalizationTheory of Relational Calculus and its Formalization
Theory of Relational Calculus and its Formalization
 
Volume computation and applications
Volume computation and applications Volume computation and applications
Volume computation and applications
 
Theory of computing
Theory of computingTheory of computing
Theory of computing
 
Regular Languages
Regular LanguagesRegular Languages
Regular Languages
 
Theory of Computation "Chapter 1, introduction"
Theory of Computation "Chapter 1, introduction"Theory of Computation "Chapter 1, introduction"
Theory of Computation "Chapter 1, introduction"
 
draft
draftdraft
draft
 
Jarrar.lecture notes.aai.2011s.ch7.p logic
Jarrar.lecture notes.aai.2011s.ch7.p logicJarrar.lecture notes.aai.2011s.ch7.p logic
Jarrar.lecture notes.aai.2011s.ch7.p logic
 
Logistic Regression(SGD)
Logistic Regression(SGD)Logistic Regression(SGD)
Logistic Regression(SGD)
 
Theory of computation Lec1
Theory of computation Lec1Theory of computation Lec1
Theory of computation Lec1
 
Pnp
PnpPnp
Pnp
 
Automata
AutomataAutomata
Automata
 
CMSC 56 | Lecture 9: Functions Representations
CMSC 56 | Lecture 9: Functions RepresentationsCMSC 56 | Lecture 9: Functions Representations
CMSC 56 | Lecture 9: Functions Representations
 
3.1 intro toautomatatheory h1
3.1 intro toautomatatheory  h13.1 intro toautomatatheory  h1
3.1 intro toautomatatheory h1
 
CommunicationComplexity1_jieren
CommunicationComplexity1_jierenCommunicationComplexity1_jieren
CommunicationComplexity1_jieren
 
Probabilistic information retrieval models & systems
Probabilistic information retrieval models & systemsProbabilistic information retrieval models & systems
Probabilistic information retrieval models & systems
 
Ddns
DdnsDdns
Ddns
 
Regular language and Regular expression
Regular language and Regular expressionRegular language and Regular expression
Regular language and Regular expression
 
Exchanging OWL 2 QL Knowledge Bases
Exchanging OWL 2 QL Knowledge BasesExchanging OWL 2 QL Knowledge Bases
Exchanging OWL 2 QL Knowledge Bases
 

Destaque

My family description
My family descriptionMy family description
My family descriptionPaula Mozu
 
Tell Me About Your Family
Tell Me About Your FamilyTell Me About Your Family
Tell Me About Your FamilySandra MP
 
Description of a room
Description of a roomDescription of a room
Description of a roomLidiaGon
 
Description of my family
Description of my familyDescription of my family
Description of my familyCatartesaura
 
Bedroom description
Bedroom descriptionBedroom description
Bedroom descriptiongvinyals
 
A description of my room
A description of my roomA description of my room
A description of my roomAnastasiaSwan78
 

Destaque (7)

Presentación1
Presentación1Presentación1
Presentación1
 
My family description
My family descriptionMy family description
My family description
 
Tell Me About Your Family
Tell Me About Your FamilyTell Me About Your Family
Tell Me About Your Family
 
Description of a room
Description of a roomDescription of a room
Description of a room
 
Description of my family
Description of my familyDescription of my family
Description of my family
 
Bedroom description
Bedroom descriptionBedroom description
Bedroom description
 
A description of my room
A description of my roomA description of my room
A description of my room
 

Semelhante a Data Complexity in EL family of Description Logics

Jarrar.lecture notes.aai.2011s.descriptionlogic
Jarrar.lecture notes.aai.2011s.descriptionlogicJarrar.lecture notes.aai.2011s.descriptionlogic
Jarrar.lecture notes.aai.2011s.descriptionlogicPalGov
 
leanCoR: lean Connection-based DL Reasoner
leanCoR: lean Connection-based DL ReasonerleanCoR: lean Connection-based DL Reasoner
leanCoR: lean Connection-based DL ReasonerAdriano Melo
 
applied mathematics methods.pdf
applied mathematics methods.pdfapplied mathematics methods.pdf
applied mathematics methods.pdfCyprianObota
 
Algorithms and Complexity: Cryptography Theory
Algorithms and Complexity: Cryptography TheoryAlgorithms and Complexity: Cryptography Theory
Algorithms and Complexity: Cryptography TheoryAlex Prut
 
Automata
AutomataAutomata
AutomataGaditek
 
Computational Complexity: Oracles and the Polynomial Hierarchy
Computational Complexity: Oracles and the Polynomial HierarchyComputational Complexity: Oracles and the Polynomial Hierarchy
Computational Complexity: Oracles and the Polynomial HierarchyAntonis Antonopoulos
 
Design and Analysis of Algorithms Exam Help
Design and Analysis of Algorithms Exam HelpDesign and Analysis of Algorithms Exam Help
Design and Analysis of Algorithms Exam HelpProgramming Exam Help
 
Jarrar.lecture notes.aai.2011s.ch8.fol.introduction
Jarrar.lecture notes.aai.2011s.ch8.fol.introductionJarrar.lecture notes.aai.2011s.ch8.fol.introduction
Jarrar.lecture notes.aai.2011s.ch8.fol.introductionPalGov
 
End semexam | Theory of Computation | Akash Anand | MTH 401A | IIT Kanpur
End semexam | Theory of Computation | Akash Anand | MTH 401A | IIT KanpurEnd semexam | Theory of Computation | Akash Anand | MTH 401A | IIT Kanpur
End semexam | Theory of Computation | Akash Anand | MTH 401A | IIT KanpurVivekananda Samiti
 
[ABDO] Logic As A Database Language
[ABDO] Logic As A Database Language[ABDO] Logic As A Database Language
[ABDO] Logic As A Database LanguageCarles Farré
 
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015Сергей Кольцов —НИУ ВШЭ —ICBDA 2015
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015rusbase
 
A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...
A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...
A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...mathsjournal
 

Semelhante a Data Complexity in EL family of Description Logics (20)

Jarrar.lecture notes.aai.2011s.descriptionlogic
Jarrar.lecture notes.aai.2011s.descriptionlogicJarrar.lecture notes.aai.2011s.descriptionlogic
Jarrar.lecture notes.aai.2011s.descriptionlogic
 
leanCoR: lean Connection-based DL Reasoner
leanCoR: lean Connection-based DL ReasonerleanCoR: lean Connection-based DL Reasoner
leanCoR: lean Connection-based DL Reasoner
 
applied mathematics methods.pdf
applied mathematics methods.pdfapplied mathematics methods.pdf
applied mathematics methods.pdf
 
Algorithms and Complexity: Cryptography Theory
Algorithms and Complexity: Cryptography TheoryAlgorithms and Complexity: Cryptography Theory
Algorithms and Complexity: Cryptography Theory
 
W.cholamjiak
W.cholamjiakW.cholamjiak
W.cholamjiak
 
Automata
AutomataAutomata
Automata
 
Computational Complexity: Oracles and the Polynomial Hierarchy
Computational Complexity: Oracles and the Polynomial HierarchyComputational Complexity: Oracles and the Polynomial Hierarchy
Computational Complexity: Oracles and the Polynomial Hierarchy
 
L16
L16L16
L16
 
C2.0 propositional logic
C2.0 propositional logicC2.0 propositional logic
C2.0 propositional logic
 
APAL2032
APAL2032APAL2032
APAL2032
 
Design and Analysis of Algorithms Exam Help
Design and Analysis of Algorithms Exam HelpDesign and Analysis of Algorithms Exam Help
Design and Analysis of Algorithms Exam Help
 
Jarrar.lecture notes.aai.2011s.ch8.fol.introduction
Jarrar.lecture notes.aai.2011s.ch8.fol.introductionJarrar.lecture notes.aai.2011s.ch8.fol.introduction
Jarrar.lecture notes.aai.2011s.ch8.fol.introduction
 
Unit 3
Unit 3Unit 3
Unit 3
 
Unit 3
Unit 3Unit 3
Unit 3
 
End semexam | Theory of Computation | Akash Anand | MTH 401A | IIT Kanpur
End semexam | Theory of Computation | Akash Anand | MTH 401A | IIT KanpurEnd semexam | Theory of Computation | Akash Anand | MTH 401A | IIT Kanpur
End semexam | Theory of Computation | Akash Anand | MTH 401A | IIT Kanpur
 
H-MLQ
H-MLQH-MLQ
H-MLQ
 
20320140501020
2032014050102020320140501020
20320140501020
 
[ABDO] Logic As A Database Language
[ABDO] Logic As A Database Language[ABDO] Logic As A Database Language
[ABDO] Logic As A Database Language
 
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015Сергей Кольцов —НИУ ВШЭ —ICBDA 2015
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015
 
A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...
A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...
A POSSIBLE RESOLUTION TO HILBERT’S FIRST PROBLEM BY APPLYING CANTOR’S DIAGONA...
 

Último

Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdfhans926745
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsEnterprise Knowledge
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Igalia
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...HostedbyConfluent
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘RTylerCroy
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Paola De la Torre
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfEnterprise Knowledge
 
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Alan Dix
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking MenDelhi Call girls
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerThousandEyes
 

Último (20)

Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
 
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 

Data Complexity in EL family of Description Logics

  • 1. Data Complexity in EL family of Description Logics Adila A. Krisnadhi (Joint work with Dr. Carsten Lutz (TU Dresden/Univ. Bremen) 2007
  • 2. Introduction Increased interest in lightweight description logics (DLs) Admits tractable reasoning in large scale ontologies. EL family: tractable ontology language; sufficient expressive power for modeling life-science ontologies DL-Lite family: tailored towards applications with massive amount of instance data Most relevant reasoning services: instance checking: decision problem asking whether an individual is an instance of a concept/class w.r.t. background knowledge base (KB) conjunctive query answering: search problem related to a generalized form of instance checking Relevant complexity measure: data complexity (measured w.r.t. the size of instance data only), rather than combined complexity (measured w.r.t. the whole input: instance data, KB schema, query)
  • 3. Objective Aim: investigate data complexity for the EL family We show data (in)tractability for wide range extensions of EL in the following. Data tractable = polynomial data complexity. Data-intractable (coNP-hard, already for instance checking): EL∀r .⊥ , EL∀r .C , ELC D + EL∃¬r .C , EL∃r .C , EL∃r ∪s.C EL≤kr , EL≥kr for some fixed integer k ≥ 0 ELkf : EL + k-functional roles (i.e., at most k successors), k > 1 (Baader,et.al.,2005): the above DLs are at least PSpace-hard (most are ExpTime-hard) regarding combined complexity. Data-tractable (for conjunctive query answering): ELI f : EL + inverse role + (1)-functional role (Baader,et.al.,2005): Regarding combined complexity, ELI is PSpace-hard and ELf is ExpTime-complete (Hustadt,et.al.,2005): ELI f is data-tractable for instance checking Data-tractability results for DL-Lite family (Calvanese, et.al., 2006)
  • 4. EL syntax & semantics Syntax: based on set of concept names NC and role names NR Set of (EL)-concepts (i.e., class in OWL): A is a concept for every A ∈ NC is a concept and if C , D are concepts, then C D, ∃r .C are concepts too for r ∈ NR Semantics: Based on an interpretation I = (∆I , ·I ) where the domain ∆I is a set of individuals and ·I maps each concept to a set of individuals in ∆I and each role to a binary relation on ∆I . AI ⊆ ∆I for each A ∈ NC r I ⊆ ∆I × ∆I for each r ∈ NR I = ∆I (C D)I = C I ∩ D I (∃r .C )I = {x | ∃y : (x, y ) ∈ r I and y ∈ C I }
  • 5. EL knowledge base (syntax) Consists of a TBox T : a set of concept definitions and/or concept inclusions . concept definition: statement otf A = C , A concept name, C concept concept inclusion: statement otf C D, C , D are concepts Two kinds of TBoxes: acyclic TBox: only allow concept definitions, LHS of all definitions are unique, has no cyclic definition general TBox: may also allow concept inclusion an ABox A: a set of concept assertions and role assertions concept assertion: expression otf A(a), A concept name, a individual name role assertion: expression otf r (a, b), r role name, a, b individual names All individual names are taken from a set NI .
  • 6. EL KB (semantics) In an interpretation I, each individual name a ∈ NI is mapped to a domain individual aI ∈ ∆I I = (∆I , ·I ) satisfies the statements/expressions: . A = C iff AI = C I C D iff C I ⊆ D I A(a) iff aI ∈ AI r (a, b) iff (aI , b I ) ∈ r I I satisfies (i.e., a model of) a TBox T and an ABox A iff it satisfies all statements and expressions in them
  • 7. Reasoning problems Satisfiability: “given a concept C and a KB (T , A), is there a model I of the KB that satisfies C , i.e., C I = ∅?” Subsumption: “given concepts C , D and a KB, is C I ⊆ D I for every model I of the KB?” KB Consistency: “given a KB, does the KB have a model?” Instance checking: “given an individual name a, a concept C and a KB K, is aI ∈ C I for every model I of the KB?” (written K |= C (a)) These reasoning tasks can (usually) be reduced to each other. Satisfiability and KB consistency in EL is trivial; subsumption and instance checking in EL is tractable, even with a larger language (EL++ ) Instance checking (and conjunctive query answering) emphasize reasoning over individuals and ABoxes, hence data complexity.
  • 8. Conjunctive query answering & entailment Conjunctive query: a set q of atoms otf C (v ) and r (u, v ); u, v variables Given an interpretation I and a mapping π that maps the set of variables in q to ∆I , I satisfies C (v ) (resp. r (u, v )) w.r.t. π iff π(v ) ∈ C I (resp. (π(u), π(v )) ∈ r I ) I satisfies q w.r.t. π (written I |=π q) iff I satisfies all atoms in q w.r.t. π I satisfies q (written I |= q) iff I |=π for some π Given knowledge base K: K |= q iff for every model of K, we have I |= q. Conjunctive query entailment: given K and q, decide whether K |= q Conjunctive query answering: given K and q, find all tuples of individual names in K such that when variables in q are properly substituted with the individual names, we have that K |= q Instance checking is a special case of conjunctive query entailment (i.e., with a single atom) Conjunctive query answering is the search problem corresponding to conjunctive query entailment.
  • 9. Data intractibility results: overview All data-intractabilty results (coNP-hardness) are for instance checking w.r.t. acyclic TBoxes. The matching upper bound (in coNP) for most results is obtained from (Grimm,et.al.,2007) for SHIQ If the TBox is empty, most cases become data tractable: since ABoxes contain only concept names, either no query contains complex concept constructor (thus trivially reduced to query answering in EL which is tractable), or can be shown directly (the ELC D case). Exception: ELkf , k ≥ 2, instance checking is coNP-complete already when the TBox is empty. Some cases distinguish whether unique name assumption (UNA) is adopted. UNA: for two individual names a, b, a = b iff aI = b I
  • 10. 2+2-SAT Data intractability: coNP-hardness results by reduction from known NP-hard problem; (Schaerf, 1993): data complexity of instance checking for EL¬A is coNP-hard, by reduction from the NP-complete problem 2+2-SAT 2+2-SAT: decide whether a given 2+2-formula is satisfiable 2+2-formula: a finite conjunction of 2+2-clause 2+2-clause: a propositional logic formula otf p1 ∨ p2 ∨ ¬n1 ∨ ¬n2 where each disjunct is a propositional letter or a truth constant 1, 0. Reduction from 2+2-SAT to (negation of) instance checking for several extensions of EL will show coNP-hardness of instance checking for the corresponding DL.
  • 11. General method for reduction Let ϕ = c0 ∧ · · · ∧ cn−1 be 2+2-SAT formula containing m propositional letters q0 , . . . , qm−1 . Let ci = pi,1 ∨ pi,2 ∨ ¬ni,1 ∨ ¬ni,2 for all i < n. Define a TBox T , an ABox A and a query concept C such that Size of A depends polynomially on size of ϕ Size of T and C are constant Show that A, T |= C (f ) (for some individual name f ) iff ϕ is satisfiable
  • 12. EL¬A (Schaerf, 1993) EL¬A = EL + atomic negation (negation only on concept names). Semantics: (¬A)I = ∆I AI Start from ϕ as defined earlier, define the following T , A and C : . T := {A = ¬A} A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )}∪ {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )} i<n C := ∃c.(∃p1 .A ∃p2 .A ∃n1 .A ∃n2 .A) where f , 1, 0, q0 , . . . , qm−1 , c0 , . . . , cn−1 are individual names and c, p1 , p2 , n1 and n2 are role names. Note: size of A is polynomial in the size of ϕ A, T |= C (f ) iff ϕ is satisfiable. (See picture) Idea: C expresses that ϕ is not satisfied, i.e., there’s a clause in which the two positive literals and the two negative literals are false. (LR): Given a model I of A, T such that f I ∈ C I , show that a / truth assignment that satisfies ϕ exists. (RL): Given a truth assignment satisfying ϕ, construct a model of A, T that does not satisfy C (f ).
  • 13. EL∀r .⊥ Essential technique from the previous proof: A and ¬A (in the TBox) partition the domain ∆I , i.e., every element in ∆I is either in AI or in (¬A)I . For EL∀r .⊥ (extension with ∀r .⊥), the partitioning concepts are ∃r . and ∀r .⊥ Semantics: (∀r .bot)I = {x | ¬∃y : (x, y ) ∈ r I } Reduction proceeds like in EL¬A using the same ABox A and query concept C but the following TBox: . . T = {A = ∃r . , A = ∀r .⊥} Again, A, T |= C (f ) iff ϕ is satisfiable.
  • 14. EL(≤kr ) , fixed k ≥ 1 EL( ≤ kr ): extension with (≤ kr ) No partitioning concepts; use covering concepts: ∃r . and (≤ kr ) I I Semantics: (≤ kr ) = {x | #{y | (x, y ) ∈ r } ≤ k} where #S denotes cardinality of the set S. Reduction: ABox A and query concept C remains the same. TBox: . . T = {A = ∃r . , A = (≤ kr )} A, T |= C (f ) iff ϕ is satisfiable (LR): From a model I of A, T that doesn’t satisfy C (f ), define a truth assignment s.t. t(qi ) = 1 implies qiI ∈ AI and t(qi ) = 0 I implies qiI ∈ A . Such a truth assignment exist (due to covering), but needs not be unique since the covering concepts need not be disjoint. However, such a truth assignment always satisfies ϕ. (RL): Given a truth assignment t, define a model I of A, T that does not satisfy C (f ). Basically, the model resembles the ABox with additional individual d, and interprets A, A and r as follows: I AI = {1} ∪ {qi | i < m and t(qi ) = 1} A = {∆I AI } r I = {(e, d) | e ∈ AI }
  • 15. EL∀r .C , EL∃¬r .C EL∀r .C : extension with ∀r .C where C is a concept Semantics: (∀r .C )I = {x | ∀y : (x, y ) ∈ r I → y ∈ C I }. Reduction for EL∀r .C uses the covering concepts: ∃r . and ∀r .X EL∃¬r .C : extension with ∃¬r .C where C is a concept Semantics: (∃¬r .C )I = {x | ∃y : (x, y ) ∈ r I and y ∈ C I }. / Reduction for EL∃¬r .C uses the covering concepts: ∃r . and ∃¬r .
  • 16. + ELC D , EL∃r .C ELC D : extension with C D; (C D)I = C I ∪ D I Reduction: use the following TBox, ABox and query concept: . . . T := {V = X Y,A = X,A = Y} A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )} ∪ {V (qi ) | i < m}∪ {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )} i<n C := ∃c.(∃p1 .A ∃p2 .A ∃n1 .A ∃n2 .A) + EL∃r .C : extension with ∃r + .C ; (∃r + .C )I = {x | ∃y : (x, y ) ∈ (r I )+ ∧ y ∈ C I } where the ’+’ indicates the transitive closure of the corresponding role name. Reduction uses the same ABox and query concept as above and . . . TBox T = {V = ∃r + .C , A = ∃r .C , A = ∃r .∃r + .C } Similar reduction can also be done for EL∃r ∪s.C .
  • 17. EL(≥kr ) without UNA, k ≥ 2 No two concepts a priori cover the domain. Hence, add more structures in the ABox. E.g. k = 3, reduction uses the same query concept C as before with . . the TBox T = {A = ∃r 4 . , A = (≥ 3r )} and ABox: A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )} ∪ {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )} i<n ∪ {r (qi , b1 ), r (qi , b2 , r (qi , b3 ), r (b1 , b2 ), r (b2 , b3 ), r (b1 , b3 )} i<m See picture Either two of b1 , b2 , b3 identify the same domain element or they do not. Hence, A and A defined in T provide the covering
  • 18. EL(≥kr ) with UNA, k ≥ 2 E.g. k = 3, reduction uses the same query concept C as before with . . . the TBox T = {V = ∃r .B, A = ∃r .(A B), A = (≥ 3r )} and ABox: A := {A(1), A(0)} ∪ {c(f , c0 ), . . . , c(f , cn−1 )} ∪ {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )} i<n ∪ {r (qi , b1 ), r (qi , b2 , V (qi ), A(b1 ), A(b2 )} i<m See picture In all models I of the KB, there is a d s.t. (qiI , d) ∈ r I for some i < m. Either d = bj (qi satisfies ∃r .(A B), or not (qi satisfies (≥ 3 r )). This reduction does not works without the UNA.
  • 19. ELkf , k ≥ 2, without UNA This DL is EL extended with (global) k-functionality a role r is globally k-functional iff every element in the domain has at most k r -successor (i.e., satisfies the GCI (≤ k r )) E.g., k = 2, reduction can be done without TBox and the following query concept and ABox: A := {r (1, e), A(e), B(e), r (0, e0 ), r (e0 , e1 ), r (e1 , e2 )} ∪ {c(f , c0 ), . . . , c(f , cn−1 )} ∪ {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )} i<n ∪ {r (qi , b1 ), r (qi , b2 ), r (qi , b3 ), r (b1 , b2 ), A(b1 ), A(b2 ), B(b3 )} i<m C := ∃c.(∃p1 .∃r 3 . ∃p2 .∃r 3 . ∃n1 .∃r .(A B) ∃n2 .∃r .(A B))
  • 20. ELkf , k ≥ 2, with UNA With UNA and without TBoxes, instance checking (and conjunctive query answering) is data-tractable. Consider the input ABox as complete description of an interpretation, check all possible matches of the query. (Taking into account possible inconsistency in the ABox). E.g., k = 3, reduction for instance checking with acyclic TBoxes: . T := {V = ∃r .B} A := {r (1, d1 ), r (1, d2 ), r (1, d3 ), s(1, d1 ), B(d1 )} ∪ {r (0, e1 ), r (0, e2 ), r (0, e3 ), s (0, e2 ), s (0, e3 ), B(d2 ), B(d3 )} ∪ {c(f , c0 ), . . . , c(f , cn−1 )} ∪ {p1 (ci , pi,1 ), p2 (ci , pi,2 ), n1 (ci , ni,1 ), n2 (ci , ni,2 )} i<n ∪ {V (qi ), r (qi , bi,1 ), r (qi , bi,2 ), r (qi , bi,3 ), s(qi , bi,1 ), s (qi , bi,2 ), s (qi , bi,3 )} i<m C := ∃c.(∃p1 .∃s .B ∃p2 .∃s .B ∃n1 .∃s.B ∃n2 .∃sB) qi satisfies either ∃s.B or ∃s .B
  • 21. Lower bounds result summary EL extension w.r.t. acyclic TBoxes w.r.t. general TBoxes EL¬A coNP-complete coNP-complete ELC D coNP-complete coNP-complete EL∀r .⊥ , EL∀r .C coNP-complete coNP-complete EL(≤kr ) , k ≥ 0 coNP-complete coNP-complete ELkf w/o UNA, k ≥ 2 coNP-complete (even w/o TBox) coNP-complete ELkf with UNA, k ≥ 2 coNP-complete (in P w/o TBox) coNP-complete EL(≥kr ) , k ≥ 2 coNP-complete coNP-complete EL∃¬r .C coNP-hard coNP-hard EL∃r ∪s.C coNP-hard coNP-hard + EL∃r .C coNP-hard coNP-hard
  • 22. ELI f : data-tractability overview ELI f : EL extended with inverse roles and functional roles inverse role: r − ; (r − )I = {(y , x) | (x, y ) ∈ r I } (globally) functional role; r is (globally) functional if it satisfies the GCI (≤ 1 r ) in ELI f , both a role r and its inverse can be declared functional. wlog. no inverse role in ABox and query concept. In this section, we consider general TBoxes. (Hustadt,et.al.,2005) Data-tractability results for Horn-SHIQ implies that instance checking for ELI f w.r.t. general TBoxes is data-tractable. Its direct proof is in my master’s thesis. We show the data-tractability result of conjunctive query answering for ELI f w.r.t. general TBoxes by giving a decision procedure for conjunctive query entailment running in polytime in the size of input ABox.
  • 23. Assumptions We assume TBoxes are in normal form, i.e., GCIs are otf A B, A1 A2 B, A ∃r .B, ∃r .A B (≤ 1 r ) Every ELI f TBox T can be converted into normal form T in polytime by introducing fresh concept name. For every ABox A and conjunctive query q not using any of the concept names that occur in T but not in T , we have A, T |= q In all atoms C (v ) in a conjunctive query q, C is a concept name (i.e., no complex concept occurs in q) Can easily be achieved if C is not a concept name: replace C (v ) in q . with A(v ) and add A = C to the TBox where A is a concept name. Conjunctive queries are connected, i.e., for all variables u, v appearing in q, there are atoms r (u0 , u1 ), . . . , r (un−1 , un ) in q s.t. u = u0 and v = un Entailment of non-connected queries can be reduced to entailment of connected queries: if q is a non connected query, then A, T |= q iff A, T |= q for all connected components q of q (Glimm,et.al.,2007)
  • 24. Algorithm Given an input TBox T , ABox A and conjunctive query q Convert T into normal form (polytime in |T |) If UNA is made, check consistency of A w.r.t. T (polytime in |A| due to (Hustadt,et.al.2005)). If inconsistent, answer “yes”. Otherwise, A must be admissible, and thus continue. If UNA is not made, convert A to make it admissible w.r.t. T by identifying individuals (polytime in |A|) Construct initial canonical structure I for T and A (polytime in |A|) Check matches of q against the above structure (polytime in |A|).
  • 25. Admissible ABox w.r.t. TBox An ABox A is admissible w.r.t. a TBox T iff the UNA is made and A is consistent w.r.t. T ; or the UNA is not made and ( (≤ 1 r )) ∈ T implies that there are no individual names a, b, c occurring in A with r (a, b), r (a, c) ∈ A and b = c. Checking whether A is admissible w.r.t. T can be done in polytime in |A|: if the UNA is made, checking consistency of A w.r.t. T is equivalent to a number of instance checking w.r.t. T which is bounded in |A| if the UNA is not made, simply identify those individual names which are r -successors of some individual for all globally functional role r ; this is doable in polytime in |A|
  • 26. Canonical structure Canonical model for A and T is the limit of the sequence of interpretations I0 , I1 , . . . defined as follows. Non-standard representation of interpretations is used: the function ·I maps every element d ∈ ∆I to a set of concept names d I , instead of every concept name A to a set of elements AI . All interpretations satisfy ∆Ii ⊆ { a, p | a ∈ Ind(A) and p ∈ ex ∗ (T )} where ex ∗ (T ): the set of all paths (sequence of existentials) for T with ε the empty path ex(T ) is the set of all existentials (concepts ∃r .A occurring in the RHS of a GCI in T ) for T Let Γ be a finite set of concept names. If A ∈ Γ and A ∃r .B ∈ T , then Γ has an ∃r .B-obligation O, where O contains B those concept names B in T such that there exists A ∈ Γ with ∃r − .A B ∈T whenever r is globally functional: those concept names B such that ∃A ∈ Γ with A ∃r .B ∈ T .
  • 27. Canonical structure (2) Start from I0 defined as: ∆I0 = { a, ε | a ∈ Ind(A)} r I0 = {( a, ε , b, ) | r (a, b) ∈ A} a, ε I0 = {A ∈ NC | A, T |= A(a)} aI0 = a, ε Construct Ii+1 from Ii : If exists, select a, p ∈ ∆Ii and an α = ∃r .A ∈ ex(T ) s.t. a, p has α-obligation O, and r is not globally functional and a, pα ∈ ∆Ii ; or / there is no b, p ∈ ∆Ii with ( a, p , b, p ) ∈ r Ii . Then do the following to get Ii+1 : add a, pα to ∆Ii if r is a role name, add ( a, p , a, pα ) to r Ii if r = s − , add ( a, pα , a, p ) to s Ii set a, pα Ii to subT (O), the closure of O under subsuming concept names w.r.t. T . Assumption: a, p is selected s.t. |p| is minimal (thus all obligations are eventually satisfied); set ex(T ) is well-ordered and the selected α is minimal for the node a, p , hence constructed canonical model is unique.
  • 28. Important lemmas The canonical model I for T and A is a model of T and A. Let I be a canonical model for T and A, and J be a model for T and A. Then there is a homomorphism h from I to J s.t.: for all individual names a, h(aI ) = aJ ; for all concept names A and all d ∈ ∆I , d ∈ AI implies h(d) ∈ AJ ; for all (possibly inverse) roles r and d, e ∈ ∆I , (d, e) ∈ r I implies (h(d), h(e)) ∈ r J Let I be a canonical model for A and T , and q a conjunctive query. Then A, T |= q iff I |= q. The above lemmas show that we can decide query entailment by looking at only the canonical model, but the problem is that the canonical model is infinite.
  • 29. Important lemmas (2) Idea: if we can show that the canonical model I satisfies q iff it satisfies q for some match π that maps all variables to elements reachable by traveling only a bounded number of role edges from some ABox individual, then we’re done. Let I be a canonical model for A and T . The initial canonical model I for A and T is obtained: ∆I = { a, p | |p| ≤ 2m + k} AI = AI ∩ ∆I r I = r I ∩ (∆I × ∆I aI = aI where m is the size of T and k is the size of q. Lemma: I |= q iff I |= q. Polytime (in |A|) construction of initial canonical model: I0 can be constructed in polytime in the size of A, obligations can computed in polytime because subsumption in ELI f is decidable and the required checks are independent of the size of A the number of elements of initial canonical model is polynomial in the size of A.
  • 30. Summary EL extension w.r.t. acyclic TBoxes w.r.t. general TBoxes EL¬A coNP-complete coNP-complete ELC D coNP-complete coNP-complete EL∀r .⊥ , EL∀r .C coNP-complete coNP-complete EL(≤kr ) , k ≥ 0 coNP-complete coNP-complete ELkf w/o UNA, k ≥ 2 coNP-complete (even w/o TBox) coNP-complete ELkf with UNA, k ≥ 2 coNP-complete (in P w/o TBox) coNP-complete EL(≥kr ) , k ≥ 2 coNP-complete coNP-complete EL∃¬r .C coNP-hard coNP-hard EL∃r ∪s.C coNP-hard coNP-hard + EL∃r .C coNP-hard coNP-hard ELI f in P P-complete For all considered extension, data-tractability can be shown iff the logic is convex regarding instances, i.e., A, T |= C (a) with C = D0 · · · Dn−1 implies A, T |= Di (a) for some i < n. (Can it be generalized?) Subtle differences such as the UNA or local vs. global functionality can have an impact on data-tractability.
  • 31. Results published in ... A. Krisnadhi. Data Complexity of Instance Checking in the EL Family of Description Logics. Master’s thesis, Technische Universit¨t a Dresden, March 2007 A. Krisnadhi & C. Lutz. Data Complexity in the EL family of DLs. In Proc.of the 20th Int. Workshop on Description Logics 2007 (DL2007), p.88–99. 2007. A. Krisnadhi & C. Lutz. Data Complexity in the EL family of Description Logics. In Proc. of the 14th Int. Conf. on Logic for Programming, AI, and Reasoning (LPAR2007), vol. 4790 of LNAI, p. 333 – 347. 2007.
  • 32. Referenced works F. Baader, S. Brandt, and C. Lutz. “Pushing the EL envelope”. In Proc. of the 19th Int. Joint Conf. on AI (IJCAI-05), pages 364–369. Morgan Kaufmann, 2005. (And its accompanying technical report). D. Calvanese, G. D. Giacomo, D. Lembo, M. Lenzerini, and R. Rosati. “Data complexity of query answering in description logics”. In Proc. of the 10th Int. Conf. on KR (KR06). AAAI Press, 2006. U. Hustadt, B. Motik, and U. Sattler. “Data complexity of reasoning in very expressive description logics”. In Proc. of the 19th Int. Joint Conf. on AI (IJCAI05), pages 466–471. Professional Book Center, 2005. B. Glimm, I. Horrocks, C. Lutz and U. Sattler. “Conjunctive Query Answering for the Description Logic SHIQ”. In Proc. of the 20th Int. Joint Conf. on AI (IJCAI-07). AAAI Press, 2007. A. Schaerf. “On the complexity of the instance checking problem in concept languages with existential quantification”. Journal of Intelligent Information Systems, 2:265–278, 1993.