<!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>
      <journal-title-group>
        <journal-title>B in front of ) as an in lusion statement:
dene a knowledge base (KB) as a tuple (T; A; D), where T is a TBox, A an
a TBox is an obje tive DBox. M satises C v D in I if We CM;I DM;I.
default theories. Let , k for</journal-title>
      </journal-title-group>
    </journal-meta>
    <article-meta>
      <pub-date>
        <year>2008</year>
      </pub-date>
      <volume>1</volume>
      <issue>6</issue>
      <fpage>2</fpage>
      <lpage>3</lpage>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>-</title>
      <p>The language of is dened in two steps: rst a on ept language, then ORALC
a modal formula language.</p>
    </sec>
    <sec id="sec-2">
      <title>B denotes that is believed, while :A: denotes that is onsidered possible,</title>
      <p>by I. We also write I (T; A) if I T and I A. For an obje tive assertion ,
every M satises BC v AC in every interpretation I.
write for and say that I satises C(a), written I C(a), if a 2 ; CI CM;I CI
or that is onsistent with ones beliefs. B is stronger than A in the sense that
Interpretation of an obje tive on ept is independent of M, hen e we may
if I for every I su h that I (T; A). A0
I A and I T if every element in the respe tive ABox and TBox is satised
we write T; A if I for every I su h that I (T; A). Similarly, T; A A0
similarly for roles, role assertions and in lusion statements. Hen e we may write
U1</p>
      <p>M2
in Conversely, any model of B’ ^ :B’ must also have the shape of M1. M1,
that ’ redu es to if ’ , where is the reexiv e transitive losure of .
of ^ and _, and ’ is identied with ’ ^ T and ’ _ F; this implies that T and
in the ourse of a binding operation, i.e. after the expand rule has been applied:
formula is semi-normal if it is of the form OA ^ for an ABox A and a possibly
The last two rules of are stri tly in ^ . These do not preserve strong C1
result of substituting every o urren e of in ’ with . Substitution is performed
hM (a)=V (a)i for a prime modal atom M (a) and V 2 f&gt;; ?g is a binding, whi h
Generalizing the rewrite system for the underlying propositional language in
For this reason we have the rules whi h regain the property, should it be lost M4
Applying the relation exhaustively before the ^ is applied guarantees a orre t
that the prime modal atom BC(a) has this property, whereas B(C(a)) do not.</p>
      <p>B(&gt;(a)) (B&gt;)(a). When there are no prime modal atoms left in O’, one may
with this property is obje tive if no prime modal atoms o ur in it. Observe
rules of the relation are based on strong equivalen es, whereas the ^ relation
of rules l r in Fig. 1, while ^ is the union of and the set of rules l ^ r.
the last two rules of have not been applied. To hara terize this we say that a C1
For formulae and , h= i is a substitution fun tion: ’h= i denotes the
extends with rules whose underlying equivalen es are merely weak. We say
wrt. if there is no formula su h that ’ . The same notation is used for
empty set of formulae of the form AC(a) and :(AC(a)) with C obje tive and
Redu tion an be performed on any subformula. A formula ’ is on normal form
stri tly on the formula level for the reason that assertions do not onsist of
exhaustively to a formula results in formulae of a form whi h ree ts that OR’
T; A 6 :C(a).
of a value for an assertion C(a) will not apply to, e.g., the assertion C t D(a)
but will apply to the equivalent formula C(a) _ D(a). The substitution fun tion
[20℄, the system in Fig. 1 onsists of two rewrite relations on formulae. The
subassertions in the sense that formulae onsist of subformulae. A substitution
rewrite pro ess.
equivalen e and are hen e not sound in all ontexts. Applying the relation
F behave as empty onjun tion and disjun tion resp. We dene to be the set
Bindings might break this property, e.g., B(BC(a))hBC(a)=&gt;(a)i = B(&gt;(a)).
apply the ollapse rules to redu e (or ollapse) O’ ^ M (a) to either O’ or F.
binds M (a) to V (a).</p>
      <p>The expand rule works by binding prime modal atoms in O’, assuming no
subformula of ’ is of the form B or A for an obje tive formula . A formula
the ^ relation. Redu tion is performed modulo ommutativity and asso iativity
a knowledge base, ea h disjun t represents a Reiter extension of ’ (when OR</p>
    </sec>
    <sec id="sec-3">
      <title>Modal Redu tion Theorem. Hen e the notion of extension makes no appeal to</title>
      <p>an essentially unique weak model. It is hen e possible, within the logi itself, to
disappears. It does, however, reappear in the side onditions of the ollapse rules.
The Modal Redu tion Theorem. For ea h = (T; A; D), there are ABoxes
the formula language, only the on ept language.
Proof. By ompleteness (Theorem 1) and soundness (Theorems 2 and 3).
toepistemi extensions (stable expansions) an be dened from the Modal Redu tion
that A F, hen e
Mathemati ian). Let D onsist of this in lusion, and let J = fBobg:
Theorem for O in the same way.
the TBox. Let T = fManager v Employeeg, and let
4 The Modal Redu tion Theorem
The rewrite system is just strong enough for establishing the Modal
ReWe translate an entire knowledge base = (T; A; D) into a formula as
foldu tion Theorem: It is sound and
Modal Redu tion Theorem.
(adapted from [9℄) is represented as BEmployee u :AManager v (Engineer t
disjun t is of the form OA for some ABox A. In fa t, ea h of these disjun ts has
tu
disjun tions of the appropriate type. It is, however, not omplete for the logi
some non-empty set of individuals J . Observe that the TBox T seemingly
This formula an be redu ed to a simpler form, depending on the ABox A and
Example 2. The ALC-default Employee : :Manager = Engineer t Mathemati ian
en odes a knowledge base, it is logi ally equivalent to a disjun tion, where ea h
= Engineer t Mathemati ian(Bob). For any A su h</p>
    </sec>
  </body>
  <back>
    <ref-list />
  </back>
</article>