See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/2275964

Adding Epistemic Operators to Concept Languages

Article · July 1999

Source: CiteSeer

CITATIONS READS
57 80

5 authors, including:

Francesco M. Donini Maurizio Lenzerini

Tuscia University - Viterbo Sapienza University of Rome
217 PUBLICATIONS 5,599 CITATIONS 419 PUBLICATIONS 22,118 CITATIONS

SEE PROFILE SEE PROFILE

Daniele Nardi Andrea Schaerf

Sapienza University of Rome University of Udine
421 PUBLICATIONS 15,945 CITATIONS 179 PUBLICATIONS 6,367 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Francesco M. Donini on 31 May 2013.

The user has requested enhancement of the downloaded file.

 Adding Epistemic Operators to Concept Languages

Francesco M. Donini, Maurizio Lenzerini Werner Nutt

Daniele Nardi, Andrea Schaerf German Research Center for
Dipartimento di Informatica e Sistemistica Arti cial Intelligence (DFKI)
Universita di Roma \La Sapienza" Stuhlsatzenhausweg 3
Via Salaria 113, I-00198 Roma, Italy D-6600 Saarbrucken, Germany

Abstract have still to be addressed in order to build principled

practical systems based on concept languages (see for
We investigate the use of epistemic operators example [7, 20]). The following issues, among others,
in the framework of concept languages (al- are considered especially important:
so called terminological languages). The re-
sults of this work have a twofold signi cance.  providing knowledge representation sys-
From the point of view of epistemic logics, tems based on concept languages with powerful
our contribution is to have identi ed an ef- and sophisticated querying capabilities;
fective procedure for the problem of answer-  making concept languages able to interface with
ing epistemic queries posed to a knowledge databases;
base expressed in the concept language ALC .
From the point of view of concept languages,  providing a formal account of several mechanisms
the most relevant aspect of our work is that generally used in implemented systems, such as
we have reconstructed in logic several com- trigger rules [2, 21] or default rules, that are not
mon features of existing knowledge represen- directly expressible in rst order logic.
tation systems. Epistemic operators provide
a highly expressive query language; allow for Recent work on data and knowledge bases exploits
the treatment of several database features, the use of epistemic operators for improving both the
such as closed world reasoning and integri- expressiveness of knowledge representation languages
ty constraints; and nally can give a formal and their associated querying facilities. In particular,
characterization of some procedural mecha- it has been argued that queries should be permitted
nisms, such as trigger rules, usually consid- to address aspects of the external world as represent-
ered in frame-based systems. ed by the knowledge base, as well as aspects of what
the knowledge base knows about the external world
(see [10, 12, 16]).
1 Introduction The need for such a distinction is evident when a
knowledge base contains incomplete information about
A substantial amount of research has been carried out individuals. Incompleteness may come in via existen-
in the last decade with the aim of providing a logical tial quanti cation: for example, a knowledge base may
reconstruction of frame-based knowledge representa- know that the individual susan has at least one male
tion languages. Much of this work has taken place friend (because it was told so), without knowing who
in the context of concept languages (also called termi- this friend is. Incompleteness may also arise from dis-
nological languages or term subsumption languages). junction: for example, a knowledge base may know
An important part of it has been concerned with both that andrea is a person, and that all persons are either
the study of reasoning techniques (e.g. subsumption male or female,1 without knowing exactly andrea's
algorithms) in concept languages, and the characteri- sex.
zation of their computational complexity. Recent re-
sults [3, 4, 5, 13, 18] indicate that the crucial properties The aim of our work is to propose a formal framework
of a rst order characterization of concept languages for adding epistemic operators to concept languages.
are now well understood, with regard to both expres- Our opinion is that such a framework is the appropri-
sive power of language constructs, and computational ate one for studying several extensions of concept lan-
complexity of reasoning. 1 Notice that this is just a piece of knowledge and not
Nevertheless, it is a common opinion that several issues an integrity constraint.
guages in order to go beyond their rst order formal- terest in a particular application, through the notions
ization, and to meet the most important requirements of concept and role. Intuitively, concepts represent
for the design of principled practical systems based the classes of objects in the domain of interest, while
on concept languages. Here we focus on the use of roles represent relationships between concepts. Com-
epistemic operators both in the query language, which plex concept expressions can be de ned by means of
allows also for the treatment of integrity constraints a number of language operators, applied to primitive
and closed-world reasoning, and in the formalization concepts and roles.
of trigger rules. The syntax and semantics of ALC are as follows. We
The basic building block of our proposal is the de - assume that two alphabets of symbols, one for prim-
nition (Section 2) of a new concept language, called itive concepts, and one for primitive roles, are given.
ALCK, obtained from ALC [18] (a powerful concept The letter A will always denote a primitive concept,
language including conjunction, disjunction, and nega- and the letter P will denote a role, which in ALC is
tion of concepts, together with existential and uni- always primitive. The concepts (denoted by the let-
versal quanti cation on roles) by adding an epistemic ters C and D) of the language ALC are built out of
operator in the spirit of [10, 12, 16]. One result of primitive concepts and primitive roles according to the
our work is the design of a technique (Section 3) syntax rule:
for answering epistemic queries expressed in ALCK. C; D ?! A j (primitive concept)
The technique is an extension of the tableaux-based >j (top)
method described in [4], which has been proved ex- ?j (bottom)
tremely useful for solving several reasoning and com- C u D j (intersection)
plexity problems in concept languages. C t D j (union)
Our study shows that epistemic operators can be very :C j (complement)
useful for the design of more powerful knowledge rep- 8P .C j (universal quanti cation)
resentation systems based on concept languages. In 9P .C (existential quanti cation):
particular, we show that epistemic operators enhance
the expressive power of query languages, without in- In the following, we use parentheses whenever we need
creasing the computational complexity of query an- to disambiguate concept expressions. For example, we
swering (Section 4). In addition we have found in- shall write (9P .D) u E to mean that the concept E is
teresting cases where the use of epistemic operators not in the scope of 9P .
allows one to express queries (not expressible in rst
order logic) that both have natural interpretations and Let
, called the domain, be a countably in nite set
are strictly less costly than their rst order counter- of symbols p1 ; p2; : : :, called parameters. An interpre-
parts. Moreover, epistemic operators make several ba- tation I is a function that maps every concept to a
sic features of database systems available within con- subset of
and every role to a subset of

, in
cept languages (Section 5), allowing for the construc- such a way that the following equations are satis ed:
tion of knowledge representation systems where part >I =

of the knowledge is expressed using database models,
with no mismatch between di erent portions of the ?I = ;
knowledge base. Speci cally, such operators can be (C u D)I = C I \ DI
used to query the knowledge base under the closed (C t D)I = C I [ DI
world assumption, to express integrity constraints over
a knowledge base, and to make concept languages at (:C )I =
n C I
least as expressive as relational database languages. (8P .C )I = fp1 2
j
We also show (Section 6) how ALCK can be used to 8p2 : (p1; p2) 2 P I ! p2 2 C I g
provide a semantical characterization of a procedural
mechanism considered in several frame-based systems, (9P .C )I = fp1 2
j
namely, trigger rules. 9p2 : (p1; p2) 2 P I ^ p2 2 C I g:
2 Epistemic Concepts and Knowledge An interpretation I is a model for a concept C if C I is
nonempty. A concept is satis able if it has a model and
Bases unsatis able otherwise. We say that C is subsumed by
We make use of the concept language ALC (see [4, 18]) D if C I  DI for every interpretation I .
to de ne a knowledge base.2 Concept languages allow Let O be an alphabet of symbols, called individuals.
one to express the knowledge about the classes of in- Syntactically, instance-of relationships are expressed
in terms of membership assertions, each one either of
2 Although we restrict our attention to ALC , our frame- the form C (a) (meaning a is an instance of C ) or
work can be applied to other languages as well. P (a; b) (meaning a is related to b by means of P ),
where a and b are individuals, C is a concept, and P are satis ed:3
is a role.
>I W =

;

Formally, the meaning of membership assertions is ?I W = ;

;

speci ed as follows. Let be a xed, injective function

from O to
(i.e. each individual is associated with AI W;
AI =
a unique parameter), and I be an interpretation. An PI W;
PI =
assertion C (a) is satis ed by I if (a) 2 C I . Similarly, (C u D)I W;
C I W \ DI W
= ; ;

an assertion P (a; b) is satis ed by I if ( (a); (b)) 2 (C t D)I W C I W [ DI W

=
P I . A nite set of membership assertions is called an
; ; ;

ALC -knowledge base. An interpretation I is a model (:C )I W

;

n CI W
= ;

of a knowledge base  if I satis es all its assertions. (8R.C )I W

;
fp1 2
j
=
 is satis able if it has a model. The set of models of
 is denoted as M().  logically implies  (written 8p2. (p1; p2) 2 RI W ! p2 2 C I W g
; ;

 j= ), where  is a membership assertion, if every (9R.C )I W = fp1 2
j

;

model in M() satis es . p2. (p1; p2) 2 RI W ^ p2 2 C I W g

9\ ; ;

In so-called terminological systems, the knowledge (KC )I W = (C J W )

base also includes an intensional part, called terminol-
; ;

ogy, expressed in terms of concept de nitions, which J2W

are usually assumed to be acyclic (i.e. in the de nition
\
(KP )I W =
;
(P J W ):
;

of concept C neither direct nor indirect reference to J2W

C itself may occur). It is well known that any reason-
ing process over knowledge bases comprising an acyclic Notice that, since the domain is xed independently
terminology can be reduced to a reasoning process over of the interpretation, it is meaningful to refer to the
a knowledge base with an empty terminology, since intersection of the extensions of a concept in di erent
one can substitute every concept name in the asser- interpretations. It follows that KC is interpreted in W
tions with the corresponding de nition (see [14] for a as the set of objects that are instances of C in every
discussion on this topic). For this reason, without loss model belonging to W . In this sense, KC represents
of generality we conceive a knowledge base as just a those objects known to be instances of C .
set of assertions.
In the standard approach, querying a knowledge base An ALCK-knowledge base T is a nite set of member-
 means asking whether C (a) (or P (a; b)) is logically ship assertions whose concepts and roles belong to the
implied by . The semantics associated with concept language ALCK. An epistemic model of T is a pair
languages is an open world semantics: the answer to a (I ; W ), where I 2 W and W is any maximal set of
query Q will be YES if Q is true in every model of , interpretations such that for each J 2 W , every asser-
NO if Q is false in every model, and UNKNOWN otherwise.
tion of T is true in (J ; W ). T is said to be satis able if
there exists an epistemic model of T , unsatis able oth-
The use of epistemic operators in the query language erwise. T logically implies an assertion  if  is true in
allows for a more sophisticated interaction with the every epistemic model of T . Note that, if  does not
system, as we will see in the next sections. Generally contain epistemic operators, it is an ALC -knowledge
speaking, we follow [16], and use K to mean that the base, and its epistemic models are the pairs (I ; W ),
system knows that is true in every model of . We such that W is the set M() of ordinary models of .
now introduce the language ALCK, which is an exten- Let T be an ALCK-knowledge base and let  be an
sion of ALC with epistemic operators and is de ned ALC -knowledge base (hence  is rst order). An
by the following syntax (where C; D denote concepts, epistemic -model of T is a pair (I ; M()) with
R denotes a role, A denotes a primitive concept and I 2 M() such that every assertion in T is true in
P a primitive role): (I ; M()). T is said to be -satis able if there exists
an epistemic -model of T , -unsatis able otherwise.
C; D ?! A j C u D j C t D j :C j Let C be an ALCK-concept, we write T j C (a) to
8R.C j 9R.C j KC mean that every epistemic -model of T satis es C (a),
R ?! P j KP: and we write  j C (a) to mean  j C (a).4 It is
easy to verify that  j C (a) if and only if the ALCK-
knowledge base  [f:C (a)g is -unsatis able. Let us
We give the semantics of ALCK by adapting the se- now introduce the notion of answer to a query.
mantics presented in [10, 12, 16] to the framework of 3 Notice that if one discards and W in the equations,
K
concept languages. An epistemic interpretation is a one obtains the standard semantics of ALC .
pair (I ; W ) where I is an interpretation and W is a 4 Since  is rst order, this means that for every model
set of interpretations such that the following equations I 2 M() the membership ( ) is true in (I M()).
C a ;
De nition 2.1 Given an ALC-knowledge base , an is to check whether the ALCK-knowledge base ob-
ALCK-concept C , and an individual a, the answer to tained by adding to  the negation of the query is
the query C (a) posed to  is YES if  j C (a), NO if -unsatis able. In this section we present a gener-
 j :C (a), and UNKNOWN otherwise. Moreover, the al method for the problem of checking whether an
answer set of C w.r.t.  is the set fa 2 O j  j ALCK-knowledge base is -satis able, where  is an
C (a)g, where O is the set of individuals appearing ALC -knowledge base.
in . Let V be a set of variables. The elements of V will be
We end this section with an example illustrating three denoted by x; y, while the elements of O [ V (called
epistemic queries. objects) will be denoted by w; z . Finally, the elements
of O will be denoted by a; b. A constraint is a syntactic
Example 2.2 Consider the ALC-knowledge base 1 structure of one of the forms:
and the following three ALCK-queries: w: C; wRz;
1 = fFRIEND(john; susan); FRIEND(john; peter);
LOVES(susan; peter); LOVES(peter; mary) where C is an ALCK-concept and R is an ALCK-role.
8LOVES.:Male(john); A constraint system is a nite set of constraints of the
Married u 9FRIEND.Male(susan); above forms.
Male(peter); :Married(mary)g:
In order to assign a meaning to constraints, we need
Query 1: 9FRIEND.Male(susan), i.e., is there a friend the following de nitions. An assignment (
) is a func-
of susan's who is male? Answer: YES. tion from O [ V into
, such that jO = , i.e. is
Query 2: 9KFRIEND.KMale(susan), i.e., is there an extension of to variables. Let (I ; W ) be an epis-
someone that is known to be both a friend of susan's temic interpretation, and let be an assignment. The
and male? Answer: NO. triple (I ; W ; ) is said to satisfy the constraint w: C
if (w) 2 C I W . Similarly, (I ; WI; W) satis es the con-
;

Query 3: 9KLOVES.:KMarried(susan), i.e., is there straint wRz if ( (w); (z )) 2 R . Let S be a con-

;

someone that is known to be loved by susan that is straint system, and let  be an ALC -knowledge base.
not known to be married? Answer: YES (peter). (I ; W ; ) is a solution of S if (I ; W ; ) satis es all of
The answer to Query 1 is justi ed by the explicit as- its constraints. S is said to be -solvable if there is a
sertion 9FRIEND.Male(susan). The answer to Query triple (I ; M(); ) that is a solution of S . If there is
2 is explained by the fact that fp j ( (susan); p) 2 no such solution, then S is said to be -unsolvable.
(KFRIEND)I W g is empty, for any model (I ; W ) of
;
Given an ALC -knowledge base , we de ne S to be
1 . In other words, the knowledge base knows that the constraint system that includes one constraint a: C
susan has a male friend, but it does not know who for each assertion C (a) of , and one constraint aRb
he is, and therefore, when taking the intersection of for each assertion R(a; b) of  (see [8]). The next
the extensions of FRIEND in the models of 1 , all the proposition shows that answering a query posed to a
pairs ( (susan); p) are ruled out. On the other hand, knowledge base  can be reduced to checking a par-
the answer to Query 3 is YES, since the knowledge ticular constraint system for -solvability.
base knows that susan loves peter, and it does not
know that peter is married, because peter is not in Proposition 3.1 Let  be an ALC-knowledge base,
(KMarried)I W . ;
C be an ALCK-concept, and a an individual. Then
 j C (a) i S [ fa: :C g is -unsolvable.
3 The Calculus for Query Answering An ALCK-concept is said to be simple if every comple-
The problem of designing methods for epistemic query ment appearing in it is either of the form :A or of the
answering is dealt with in [10, 16]. In [16], a procedure form :KC , where C is simple. Every ALCK-concept
is presented that is sound and complete if the query can be rewritten in linear time into an equivalent sim-
satis es some syntactic constraints. However, not all ple concept. In the rest of the paper we assume to deal
epistemic concepts belonging to ALCK satisfy such only with simple concepts.
constraints (for example, the formula corresponding to We say that wRz holds in a constraint system S if
9P .:KC (a) is not admissible in [16]). On the other either
hand, the method proposed in [10] has been conceived
within a more general framework, and its specializa- 1. R is P , and wPz 2 S , or
tion to the case of concept languages does not provide 2. R is KP , w; z 2 O, and wPz 2 S .
an e ective procedure. It follows that none of these
approaches can be directly applied in our setting. When checking the satis ability of a constraint sys-
As we said in Section 2, one way to answer epis- tem, the following set of completion rules are used (S
temic queries posed to an ALC -knowledge base  denotes a constraint system):
 1. S !u fw: C1; w: C2g [ S 4. S contains a constraint of the form a: :KC , and
if w: C1 u C2 is in S , and w: C1 and every completion of S [ fa: :C g contains a -
w: C2 are not both in S ; clash;
2. S !t fw: Dg [ S 5. S contains a constraint of the form aKPb, and
if w: C1 t C2 is in S , neither w: C1 aPb 62 S ;
nor w: C2 is in S , 6. S contains x: :KC , x: KC , xKPy, or wKPx,
and D = C1 or D = C2; and for each a 2 O [ fg, every completion of
3. S !9 fwRx; x: C g [ S S [x=a] contains a -clash, where  is an individ-
S

if w: 9R.C is in S , there is no z such ual in O n O .S

that both wRz and z : C are in S

and x is a new variable; Proposition 3.4 Let  be an ALC-knowledge base,
4. S !8 fz : C g [ S and S be a constraint system. Then S is -solvable
if and only if there exists at least one completion of S
if w: 8R.C is in S , wRz holds in S , that contains no -clash.
and z : C is not in S .
Proposition 3.2 Let  be an ALC-knowledge base, Proof. The proof is by induction on the number k of
and S; S 0 be two constraint systems. Then:
occurrences of the epistemic operator in the constraint
system. If k = 0, then the claim trivially follows from
1. If S 0 is obtained from S by application of one of the results in [1]. If k > 0, then let H1 be the following
rules !u , !9, !8, then S is -solvable if and induction hypothesis: any complete constraint system
only if S 0 is -solvable. with h < k occurrences of the epistemic operator is
-solvable if and only if it contains no -clash. Let S
2. If S 0 is obtained from S by application of the !t - be a complete constraint system with k occurrences of
rule, then S is -solvable if S 0 is -solvable. Con- the epistemic operator. We must show that
versely, if S is -solvable and the !t -rule is ap-
plicable to S , then it can be applied in such a way 1. if S contains no -clash, then it is -solvable, and
that it yields a -solvable constraint system. 2. if S is -solvable, then it contains no -clash.
A constraint system is said to be complete if no rule Proof of (1). Suppose S contains no -clash. We con-
is applicable to it. Any complete constraint system struct a triple (I ; M(); ). First of all, is de ned
obtained from a constraint system S by applying the by:
above rules is called a completion of S . Notice that,
due to the presence of the !t -rule, more than one  (w) = (w) if w is an individual;
completion can be obtained starting from a constraint  (x) = (a) if w if x is involved in at least one
system. epistemic constraint in S (i.e., such that x: :KC ,
We denote by O the set of individuals appearing in x: KC , xKPy, or wKPx is in S ), and a is any
a constraint system S . Moreover, if a is an individual,
S
individual in O [ fg such that S [x=a] does not
S

then we denote by S [x=a] the constraint system ob- contain any -clash (there must be such an a,
tained from S by substituting every occurrence of the otherwise S would contain a clash of type 6);
variable x with the individual a.  (x) = p if x is not involved in any epistemic con-
We now introduce the notion of -clash, which intu- straint, and p is any parameter that is assigned by
itively corresponds to an explicit contradiction in con- neither to an individual in O nor to a variable
y in S such that y 6= x.
S

straint systems. Then, we exploit such a notion in

order to derive a sucient and necessary condition for Secondly, I is de ned as follows:
a constraint system to be -solvable.
De nition 3.3 A constraint system S contains a -  for every primitive concept A and every parameter
clash if at least one of the following conditions holds: p, let p 2 AI if and only if there is a w such that
(w) = p, and w: A is in S ;
1. S contains a constraint of the form w: ?;  for every primitive role P andI every pair p1; p2 of
parameters, let (p1; p2) 2 P if and only if there
2. S contains two constraints of the form are w; z such that (w) = p1, (z ) = p2, and wPz
w: A; w: :A; is in S .
3. S contains a constraint of the form a: KC , and  the interpretations of complex concepts and roles
there is at least one completion of S [ fa: :C g are derived from the semantic equations given in
without -clashes; Section 2.
We show that (I ; M(); ) satis es every constraint In conclusion, we have shown that the triple
in S , i.e. S is -solvable. (I ; M(); ) is a solution of S , and therefore S is -
Consider any constraint of the form wPz . By the solvable.
construction of and I , it is easy to verify that Proof of (2). Suppose S contains a -clash. We now
( (w); (z )) 2 P I M().
;
consider each type of clash in turn, and show that if S
Consider any constraint of the form w: C . We proceed contains a clash of that type, then it is -unsolvable.
by a secondary induction on the form of C , considering 1, 2. If S contains a clash of type 1 or 2, then it is
as base cases >, A, :A, KC , and :KC . clearly -unsolvable.
If C is of one of the forms >, A, :A, then it follows 3. Suppose S contains a constraint of the form a: KC ,
by the construction of and I that (w) 2 C I M().
;
and there is at least one completion of S [ fa: :C g
Consider any constraint of the form a: :KC . Since without -clash. By the induction hypothesis H1
S does not contain any -clash, there is at least and by Proposition 3.2, S [ fa: :C g is -solvable,
one completion of S [ fa: :C g which does not con- i.e. there is a triple (I ; M(); ) that satis es all
tain any -clash. Since the number of occurrences of the constraints of S [ fa: :C g, and in particular
the epistemic operator in S [ fa: :C g is less than a: :C . Therefore,
T (a) 62 C I M(), which implies
;

k, by the induction hypothesis H1 and by Proposi- that (a) 62 J2M() C J M(). It follows that the
;

tion 3.2, S [ fa: :C g is -solvable, which means constraint a: KC cannot be satis ed by any triple
that there is a model J of , such that (a) 2 (I ; M(); ), and therefore S is -unsolvable.
(:C )J M(), andT hence (a) 62 C J M(). Since J 2
; ;
4, 5. If S contains a -clash of the form a: :KC or of
M(), (a) 62 J2M() C J M() , hence, by de ni-
;
the form aKPb, then we can proceed analogously to
tion of (KC )I M(), (a) 62 (KC )I M(). Therefore,
; ; case 3.
(I ; M(); ) satis es a: :KC . 6.1. Suppose S contains a constraint of the form
Consider any constraint of the form a: KC . Since x: :KC , and for each a 2 O [ fg, S [x=a] contains a
-clash, where  is an individual in O not appearing in
S

S does not contain any -clash, every completion of

S [ fa: :C g contains a -clash. Since the number of S . This means that for each a 2 O [fg, either every
completion of S [fa: :C g contains a -clash, or every
S

occurrences of the epistemic operator in S [ fa: :C g

is less than k, by the induction hypothesis H1 and completion of S [x=a] n fa: :KC g contains a -clash.
by Proposition 3.2, S [ fa: :C g is unsolvable, which By the induction hypothesis H1, this implies that for
means that forT every model J of , (a) 2 C J M(),
; each a 2 O [fg, either S [fa: :C g is -unsolvable,
or S [x=a] n fa: KC g is -unsolvable. We show that S
S

i.e. (a) 2 J 2M() C J M() , and hence (a) 2

;

cannot be satis ed by any triple (I ; M(); ).

(KC )I M(). Therefore, (I ; M(); ) satis es a: KC .
;

Let us assume that the triple (I ; M(); ) is a so-

Consider any constraint of the form x: KC . Let a be lution of S , with (x) = p, and there is a b 2 O
the individual in O [ fg such that (x) = (a)
S such that (b) = p. Obviously, (I ; M(); ) is a so-
S

(notice that by the construction of , S [x=a] does lution of S [x=b] too, which means both that the con-
not contain any -clash). Since S [x=a] does not con- straint b: :KC is satis ed by (I ; M(); )|i.e. that
tain any -clash, every completion of S [ fa: :C g S [ fb: :C g is -solvable|and that every constraint
contains a clash. Recall from theTprevious case that in S [x=b] n fb: :KC g is satis ed by (I ; M(); )|i.e.
the last condition implies (a) 2 J2M() C J M(),
;
S [x=b] n fb: :KC g is -solvable|which is a contradic-
and hence (a) 2 (KC )I M(). It follows that
;
tion.
(x) 2 (KC )I M(), that is, (I ; M(); ) satis es
;
Let us assume that the triple (I ; M(); ) is a solution
x: KC . The other forms of epistemic constraints, of S , with (x) = p, and there is no b 2 O such that
namely x: :KC , xKPy, and wKPx can be treated S
(b) = p. It is possible to show that in this case there is
analogously. a solution (I 0 ; M() ; 0) of S0 such that 0 (x) = ().
Now consider the following induction hypotesis H2: 0
It follows that (I ; M(); ) is a solution of S [x=],
for every proper subconcept D of a concept C , every too, which is again a contradiction as in the previous
constraint z : D is satis ed by (I ; M(); ). case.
Suppose w: C is in S and C has the form D u E . 6.2, 6.3, 6.4. If S contains a -clash of the form x: KC ,
Since S is complete, both w: D and w: E are in S . xKPy, or wKPx, then we can proceed analogously to
By the induction hypothesis H2, (I ; M(); ) satis es case 6.1.
both constraints, and therefore, (I ; M(); ) satis es
w: D u E too. The remaining forms of constraints, The results reported in [1] show that it is decidable
namely, E t D; 9R.D, and 8R.D, can be treated anal- whether a constraint system that does not include any
ogously. epistemic constraint is -solvable. In particular, it
is shown that the number of completions obtainable the presence of the K operator, the satis ability of
from such a system is nite. Based on this fact, one constraint systems obtained from the precompletion
can easily prove that the number of completions of an cannot be done by considering the constraints on each
ALCK-constraint system is also nite. Observe that, individual separately. For example, the satis ability
in order to decide whether a complete constraint sys- of a constraint of the form a: 9KP .KC depends on
tem S has a -clash or not, a nite number of - the assertions on other individuals in the knowledge
satis ability checks suces, each involving a constraint base. For instance, its satis ability follows from the
system whose number of epistemic constraints is less two assertions: P (a; b) and C (b).
than the number in S . Therefore, one can show by In the calculus shown in Section 3, this possibility is
induction that the above rules provide us with an al- taken into account by the condition for a -clash of
gorithm for checking an ALCK-constraint system for type 6. In fact, getting back to the above example, the
-solvability. constraint a: 9KP .KC is processed creating the two
Proposition 3.5 Let  be an ALC-knowledge base. constraints aKPx and x: KC , and checking whether x
Then it is decidable whether an ALCK constraint sys- can be substituted with an individual without running
tem S is -solvable or not. into a clash.
Nevertheless, answering ALCK-queries can be done in
This in turn implies that we have an e ective method PSPACE. In fact, it is still possible to devise an algo-
both for checking whether  j C (a), and for com- rithm that uses polynomial space by keeping a table
puting the answer set of C w.r.t. . The next section with the information concerning substitutions. Such
discusses in more detail the computational complexity a table has a boolean value for every subconcept of 
of the method. and of the query C , and every individual in O . The
value of an entry (D; a) is true whenever the comple-
tion process of S [ fa: :C g would add the constraint
4 Complexity and Expressiveness a: D as the result of a substitution [x=a] in a constraint
x: D. Notice that a table represents the set of con-
In this section we rst address the computational com- straints on individuals that are common to all traces
plexity of answering ALCK-queries. Then we discuss of a completion.
how the epistemic operator can be used in the for- The algorithm generates a table and then considers the
mulation of interesting queries to an ALC -knowledge constraint system S [fa: :C g[ T , where T is the set
base. of constraints represented by the table. At this point
Using a technique based on constraint systems, both the algorithm performs the same two steps of the algo-
satis ability and subsumption of ALC -concepts are rithm for ALC , but for a more complex veri cation of
are proved in [18] to be PSPACE-complete problems. clashes in the traces, where all the six types described
That paper also presents a linear space algorithm for in the previous section are considered. Clashes of type
these problems. The basic idea behind the algorithm 1,2,5 are found by inspection. Clashes of type 3,4 re-
is that, although the whole constraint system involved quire a recursive call to the algorithm. Finally in the
in the computation may have exponential size, it is check of clashes of type 6, substitutions which intro-
possible to keep track of only a polynomial part of it duce constraints not already present in the table are
at a time. These parts, called traces, are mutually in- disallowed. If there are no allowed substitutions lead-
dependent, and can be checked for a clash separately. ing to a trace with no -clash, the table is discarded
In [1] a PSPACE algorithm for checking the satis abil- and the algorithm restarts with a new one. If no ta-
ity of an ALC -knowledge base  is given, which again ble leads to the construction of a completion with no
is based on the constraint system calculus. Since for -clash, the constraint system S [ fa: :C g is not -
any ALC -concept C we have  j= C (a) if and only solvable.
if the ALC -knowledge base  [ f:C (a)g is unsatis - The size of the table is polynomially bounded by
able, it follows that answering ALC -queries to an ALC - (jj + jC j)  jOj, where jj, jC j and jO j denote the
knowledge base can be done in polynomial space, too. size of , C and O, respectively. Since each recur-
The algorithm for deciding the satis ability of ALC sive call requires a new table but reduces the number
knowledge bases consists of two steps: in the rst one, of K operators, the number of tables that are simul-
which we call precompletion, all the information re- taneously needed is bounded by the nesting of the K
garding each individual is collected in a single con- operator. Therefore, we can conclude that adding the
straint; in the second step these constraints are sep- K operator to the query language does not change the
arately tested for satis ability using the linear-space complexity class of query answering.
algorithm for concept satis ability given in [18]. Proposition 4.1 Let  be an ALC-knowledge base,
However, when the query is an ALCK-concept, the C be an ALCK concept, and a be an individual. Then
above method is no longer valid. In fact, because of checking whether  j C (a) is PSPACE-complete.
The analysis of the query answering algorithm allows that 1 6j= Married u 9LOVES.:Married(peter), and
us to make an interesting comparison between reason- since 1 6j= 9LOVES.:Married(susan), it follows that
ing with epistemic concepts and reasoning with con- 1 6j= Married u 9LOVES.:Married(susan).
cepts involving collections of individuals. A construct Reasoning in this way would lead to the answer NO.
for collection of individuals, sometimes called \ONE- On the contrary, the correct answer to the query
OF," has been considered in [2] and [9]. If a1; : : :; a is YES, and in order to nd it, one needs to rea-
are individuals, then the concept fa1; : : :; a g is inter-
n

n
preted as the set consisting of the elements denoted son by case analysis. In fact, the query asks if
by a1; : : :; a . Now, the intuition is that a concept in every model M of 1 there is an individual,
of the form KC can be considered equivalent to the
n
say z , such that FRIEND(john; z ), Married(z ), and
concept fa1; : : :; a g, where a1 ; : : :; a are exactly the 9LOVES.:Married(z ) are true in M . Obviously, in
individuals for which  j= C (a ) holds. Indeed, we
n n
every model M of 1 , either Married(peter) or
have proved that if L is any sublanguage of ALC , LK is
i
:Married(peter) is true. In the rst case, it is easy to
the language obtained from L by adding the constructs see that z is simply peter (and the unmarried person
KC , :KC , and KP , and LO is the language obtained he loves is mary), while in the second case z is susan
from L by adding the constructs fa1 ; : : :; a g and (and the unmarried person she loves is just peter).
:fa1; : : :; a g for concepts, and f(a1; b1); : : :; (a ; b )g
n
Therefore, such an individual z exists in every model
for roles, then the two problems of querying an L-
n n n
of 1 , and the query gets the answer YES.
knowledge base using LK-concepts and using LO- Note that, even if none of the individuals related to the
concepts are in the same complexity class. In particu- individual john through the role FRIEND is in the con-
lar, Proposition 4.1 implies that reasoning in ALCO is dition requested by the query, it happens that the com-
PSPACE-complete, which represents a new complexi- bination of the assertions on the individuals (susan
ty result for concept languages. and peter) in the knowledge base is such that in ev-
ery model either one or the other is in that condition.
The previous considerations allow us to conclude that On the other hand, Query 5 asks whether there
the use of epistemic operators does not substantially is an individual, say z , such that in every model
increase the complexity in query answering. This is M 2 M(1 ), both FRIEND(john; z ) and Married u
extremely interesting, especially in light of the greater 9LOVES.:Married(z ) are true. It is easy to see that
expressive power gained (see also next section). In the no such z exists in 1 , and therefore the answer is NO.
rest of this section we show how the use of the epis- Query 6 is similar to Query 5, but the concept
temic operator allows us to express queries that have :Married is replaced with the concept :KMarried.
both natural interpretation and good computational In this case, since :KMarried(peter) holds in 1, we
properties. In particular, we have found that there have that the above mentioned individual z exists. In
are situations where the use of the epistemic operator particular, it is susan, and therefore the answer is YES.
helps lessening the computational complexity of query
answering. Queries 4,5 and 6 show how the K operator is able to
Consider the following three queries posed to the ALC - modify the semantics of a query (note that, apart from
knowledge base 1 of Example 2.2: the presence of K, the queries are identical).
Query 4: Queries 4, 5, and 6 show how to use the K operator to
9FRIEND.(Married u 9LOVES.:Married)(john). modify the semantics of a query (note that they dif-
Answer: YES. fer only in the occurrences of the K operator): Query
4 respects the standard rst order semantics of con-
Query 5: cept languages, Query 5 uses an intuitionistic seman-
9KFRIEND.K(Married u 9LOVES.:Married)(john). tics which rules out the reasoning by case analysis,
Answer: NO. and nally, Query 6 makes use of a sort of negation as
Query 6: failure.
9KFRIEND.K(Married u 9LOVES.:KMarried)(john). As another example of this fact consider the following
Answer: YES. pair of queries:
Query 4 asks whether john has a married friend Query 7: (9LOVES.:Male) t (:9LOVES.>)(john).
who loves an unmarried person. At rst glance, Answer: YES.
since susan and peter are the only known friends of Query 8: K(9LOVES.:Male) t K(:9LOVES.>)(john).
john, it seems that the answer to the query is to be
found by checking whether either 1 j= Married u Answer: NO.
9LOVES.:Married(susan) or 1 j= Married u The di erence between Query 7 and 8 is similar to
9LOVES.:Married(peter). the di erence between Query 4 and 5. Query 7 asks
Since 1 6j= Married(peter), we have in particular whether in every model M 2 M(1) the assertion
(9LOVES.:Male) t (:9LOVES.>)(john) is true. 5 Closed World Reasoning, Relational
Again, reasoning by case analysis, one realizes that Databases, and Integrity
such a query gets the answer YES: indeed, due to the Constraints
assertion 8LOVES.:Male(john) in 1, for every mod-
el M of 1 , either 9LOVES.:Male(john) is true in The reason for the open world semantics of concept
M , or :9LOVES.>(john) is true in M . On the oth- languages is that they are generally used in appli-
er hand, Query 8 asks whether it is the case that cations where incomplete information have to be ac-
9LOVES.:Male(john) is true in every model M 2 counted for. For example, even if all the known friends
M(1 ), or whether :9LOVES.>(john) is true in every of John are male, one does not want to conclude that
model M 2 M(1 ). Therefore, the answer to Query 8 all friends of John are male.
is NO.
On the other hand, there are situations where it is
The above examples (Queries 5 and 8) show that the natural to query a knowledge base under the closed
use of K may allow us to express queries whose an- world assumption. It is important to note the di er-
swer does not require reasoning by case analysis. In ence between assigning a closed world semantics to the
[6], we analyze in detail the situations where this kind knowledge base and allowing one to pose queries un-
of reasoning makes deductions in concept languages der the assumption that part of the knowledge base is
problematic from the computational point of view. complete.
For example, we prove that the problem of checking
if  j= C (a), when  is an AL-knowledge base and C We argue that the use of epistemic operators as de-
is an ALE -concept, is coNP-hard w.r.t. the size of .5 scribed in the previous sections is a natural way to
achieve such a exible way of interacting with the
On the other hand, we have the following result: knowledge base. Referring to the knowledge base 1
of Example 2.2, consider the following examples:
Proposition 4.2 Let  be an AL-knowledge base, a Query 9: 8FRIEND.9LOVES.>(john).
be an individual, and C be an ALEK-concept where Answer: UNKNOWN.
the only quali ed existential quanti cations are of the Query 10: 8KFRIEND.K9LOVES.>(john).
form 9KR.KE . Then checking whether  j C (a) can Answer: YES.
be done in polynomial time w.r.t. the size of .
Query 9 gets the answer UNKNOWN because there is a
model of 1 where john is related to an individual
Notice that if C is an ALE -concept, and (C ) is the z and z is not an instance of the concept 9LOVES.>.
ALEK-concept obtained from C by replacing every oc- On the other hand, the reading of Query 10 is as fol-
currence of the construct 9R.E with 9KR.KE , then lows: is it true that for every individual z known to be
 j (C )(a) implies  j= C (a). >From this obser- related to john through the role FRIEND, it is known
vation and the above theorem we can derive an inter- that 9LOVES.>(z ) holds 1? It is easy to see that the
esting property: the algorithm for checking whether answer to this query is YES.
 j (C )(a) is a tractable, sound and incomplete The above example shows that the use of K allows one
procedure for checking whether  j= C (a). to pose queries to a knowledge base  under the as-
Following the above idea, we can use the K operator sumption that  has complete knowledge on a certain
to build sound and incomplete query answering pro- individual a and a certain role P (john and FRIEND in
cedures for any concept language. Given a knowledge the example), i.e. under the assumption that for every
base  and a query C , both expressed in a language pair (a; b) such that  6j= P (a; b), P (a; b) is false in
L, the query can be replaced by a suitable concept . Notice that this is not the same as assuming that
L (C ) in the language LK such that  j L (C )(a)
the knowledge about every role is complete, like for
implies  j= C (a) and deciding  j L (C )(a) can example in [13].
be done more eciently w.r.t. the size of . The ad- In [11] the problem of translating Circumscription into
vantage compared to other incomplete procedures is epistemic theories is discussed. We show here that our
the precise semantical characterization of the source query language allows us to achieve at least the expres-
of incompleteness. sive power of the (naive) Closed World Assumption
(CWA) (see [15]). We do so by considering knowledge
bases expressed in the simple language AL0 , whose
concepts are formed according to the rule:
5 ALE is the sublanguage of ALC that consists of all C; D ?! A j :A j C u D j 8R.C:
simple concepts which do not contain the union constructor
\t", whereas AL consists of all ALE concepts that only More complex languages and more powerful forms of
contain existential quanti cations of the form 9R.>. closed world reasoning (e.g. Careful CWA) require a
more sophisticated treatment, which is outside the method described in Section 3 is needed in order to
scope of this paper. evaluate ALCK+ -queries.
Let  be an AL0 -knowledge base, C be any ALC - ALCK+ -queries constitute a superset of relational
concept, and a be an individual.  entails C (a) under calculus queries [19]. Indeed, let us denote with
the CWA, written  j=CWA C (a), if C (a) is true in rdb() the relational database obtained from the AL0-
every minimal model of . knowledge base  as follows. First, for every primitive
Given a simple ALC -concept C , we de ne the ALCK- concept A in , we introduce a relation rel(A) of arity
concept C as follows: 1, and for every role R in , we introduce a relation
rel (R) of arity 2. Second, let rel (A) consist of the tu-
1. A = KA ples hai such that a: A is in the completion S ,6 and
let rel (R) consist of the tuples ha; bi such that aRb is
2. :A = :KA in S . It is possible to show that, for every satis able
3. (C u D) = KC u KD AL0-knowledge base , and for any relational+ calculus
expression r over rdb(), there is an ALCK -query q
4. (C t D) = KC t KD over  such that the set of tuples in the answer to
5. (9P .C ) = 9KP .C r is equal to the answer set of q. This proves that
ALCK+ has at7least the expressive power of the rela-
6. (8P .C ) = 8KP .C . tional calculus.
Proposition 5.1 Let  be an AL0-knowledge base, Another interesting feature of ALCK+ -queries is that
C be an ALC -concept, and a be an individual. Then they provide a formal setting for extracting informa-
 j=CWA C (a) if and only if  j C (a). Moreover, tion from a knowledge base where part of the knowl-
checking if  j C (a) can be done in polynomial time.
edge is expressed in a concept language, and part is
stored in a relational database.
One interesting consequence of the above result is A further aspect that is usually considered in databas-
that epistemic queries allow us to achieve the expres- es but not in concept languages, is that of integrity
sive power of relational query languages. Let us call constraints, which are sentences specifying the set of
ALCK+ the language whose syntax is as follows (C , admissible database states. In [16] it is argued that
D denote concepts, R, Q denote roles, and a denotes integrity constraints are naturally viewed as epistemic
an individual): sentences specifying what the knowledge base is sup-
posed to know about a particular aspect of the world,
C; D ?! A j fag j C u D j C t D j :C j 8R.C j rather than a direct property of the world.
9R.C j KC For example, if the we want to rule out those knowl-
R ?! P j Q edge bases which are uncertain about the sex of every
Q; Q1 ?! KP j Q u Q1 j Q t Q1 j Q  Q1 j :Q j person who is known to be student, we cannot simply
Q j Q?1 ; state that every student has a sex. Rather, we need to
specify that for every known student a, the knowledge
where fag denotes the concept whose extension is con- base either knows that a is a male, or knows that a is
stituted by the single individual a (see Section 4),  a female.
denotes role concatenation, Q denotes the transitive In order to represent this idea in our setting, we pro-
closure of Q, and Q?1 denotes the inverse of Q. pose to model integrity constraints as ALCK concepts.
An ALCK+ -query over a knowledge base  is an ex-
pression of the form De nition 5.2 An integrity constraint is an ALCK-
concept. Given an ALC -knowledge base  and an in-
hx1 ; : : :; x j i
n tegrity constraint C ,  satis es C if for every individ-
where is a rst order formula whose free variables ual a 2 O ,  j C (a). If IC = fC1; : : :; C g is a set
n

are among x1; : : :; x , whose constants are symbols of integrity constraints, then  is said to be legal with
in O , and whose atomic formulas are of+ the form
n
respect to IC if  satis es C for each i 2 f1; : : :; ng.
i

KC (t) or KQ(s;+ t), where C is an ALCK -concept, For example, the previous integrity constraint can be
Q is an ALCK -role or is the predicate symbol =, expressed as:
and s; t are either constants or+ variables. The an-
swer set of the above ALCK -query is the set of (:KStudent) t (KMale t K:Male):
tuples ha1 ; : : :; a i, where each a 2 O, such that
 j [x1 =a1; : : :; x =a ]. We are using here a straight-
n i
6 Note that, due to the language constructs of AL0 ,
forward extension of the j relation de ned in Section 2
n n

there exists exactly one completion of  .

S
to the case where the right hand side argument is a 7 ALCK+ is actually more expressive, since transitive
formula. Obviously, a corresponding extension of the closures cannot be expressed in the relational calculus [19].
Notice that integrity constraints satisfaction (i.e. One can verify that for every epistemic model (I ; W )
checking whether a knowledge base is legal w.r.t. a set of , we have (susan) 2 (8EATS.JunkFood)I W ;

of integrity constraints) can be easily realized through and (chips) 2 JunkFoodI W , i.e., both the as-
;

the calculus presented in Section 3. sertion 8EATS.JunkFood(susan) and JunkFood(chips)

are logical consequences of , as one would expect.
6 Formalizing Trigger Rules Notice that, since trigger rules are used only in one
direction, they are not expressible by logical implica-
Up to now we have considered only knowledge bases tions. Indeed, our formalization with the epistemic op-
constituted of assertions on individuals. Many practi- erator correctly captures this intuition. Consider for
cal systems for building knowledge based applications, instance the knowledge base 0 = hf:B (a)g; fKA )
such as CLASSIC [2], or CLASP [21], provide an ad- KBgi, and observe that there exists an epistemic mod-
ditional mechanism, called trigger rules, for specifying el (I ; W ) of 0 such that (a) 62 :AI . Therefore,
the knowledge base. Roughly speaking, trigger rules :A(a) is not a logical consequence of 0.
have the form: \if an individual is proved to be an
instance of C , then derive that it is also an instance We now turn our attention to the problem of deciding
of D," where C , D are concepts (see [2]). The behav- whether the speci cation h; i of a knowledge base
ior of trigger rules is usually described in terms of a is satis able or not. In order to provide a calculus for
forward reasoning process that adds to the knowledge such a problem, we add to rules 1{4 of Section 3 a new
base the assertion D(a) whenever C (a) is proved to propagation rule, dealing with the sentences in . The
hold. Let us call procedural extension of  w.r.t. to new propagation rule is as follows:
? the knowledge base resulting from such a forward 5. S !) fa: Dg [ S
reasoning process on a set of assertions  and a set of
trigger rules ?. if (KC ) KD) is in , a: D is not
Trigger rules in the context of frame-based systems in S , and every completion of S [
are often de ned informally. Attempts to precisely fa: :C g contains a -clash.
capture the meaning of such rules are based either on The above rule, together with rules 1{4 of Section 3,
viewing them as knowledge base updates (see for ex- provides a sound and complete calculus for the satis -
ample the TELL operation of [10]), or on ad hoc se- ability of h; i. Indeed, it can be shown that h; i
mantics (see [17]). is satis able if and only if there exists at least one com-
Our aim in this section is to show that trigger rules pletion of h; i without -clash (note that the only
can be nicely formalized in our setting. To this end, possible -clashes in this cases are either of type 1 or
we regard the speci cation  of a knowledge base as of type 2). Moreover, it can be shown that the calcu-
constituted by a pair h; i, where  is a set of mem- lus can be turned into an algorithm that again works
bership assertions in ALC , and is a set of epistemic in PSPACE.
sentences (representing trigger rules), each one of the In addition to the above method for checking the sat-
form KC ) KD, where C and D are ALC -concepts. is ability of a knowledge base h; i, where is a set
An epistemic interpretation (I ; W ) satis es the sen- of trigger rules, a complete account of the procedural
tence KC ) KD if (KC )I W  (KD)I W . Extending
; ;
extension requires a method for the construction of a
the de nition given in Section 2, an epistemic model new knowledge base as a result of the application of a
of h; i is a pair (I ; W ), where I 2 W and W is any set of trigger rules. To this purpose it is possible to
maximal subset of M() such that for each J 2 W , devise a re ned version of the above rule, which in-
(J ; W ) satis es every sentence in . crementally adds the consequences of trigger rules to
Intuitively, the set of epistemic sentences restricts the knowledge base . This modi cation is straight-
the set of models of  to the maximal subsets that forward in principle and we do not present it here for
satisfy every sentence in . Because of the form of such lack of space.
sentences, it can be shown that there exists only one
maximal subset W of M() such that for all J 2 W ,
(J ; W ) satis es every assertion in h; i. Moreover,
7 Conclusions
it can be shown that such W coincides with the set In the present paper we have presented a framework
of models of the procedural extension of  w.r.t. the for adding epistemic operators to concept languages.
trigger rules represented by . By enriching concept languages with an epistemic op-
erator, to distinguish what is known to the knowledge
Example 6.1 Consider the ALC-knowledge base  = base as opposed to what is true in the external world,
h; i (inspired by an example in [2]): we have been able to formally characterize several as-
 = fEATS(susan; chips); Student(susan)g pects of frame-based systems.
= fKStudent ) K(8EATS.JunkFood)g: In particular, we have considered the use of the epis-
 temic operator in the query language, and shown that representation services. Arti cial Intelligence 48
this richer query language may be used to reduce the (1991), 261{297.
complexity of reasoning within knowledge bases. In [8] Hollunder, B. Hybrid inferences in KL-ONE-
addition, by virtue of the epistemic operator one can based knowledge representation systems. In Proc.
naturally specify forms of closed-world reasoning in the of the German Workshop on Arti cial Intelli-
queries, as well as integrity constraints, and compare gence (1990), Springer-Verlag.
the query language with relational algebra. Finally, we
have shown how to formalize procedural mechanisms [9] Lenzerini, M., and Schaerf, A. Querying
such as trigger rules. concept-based knowledge bases. In Proc. of the
Workshop on Processing Declarative Knowledge,
However, several aspects of frame-based systems still PDK-91 (1991), Lecture Notes in Arti cial Intel-
remain to be explored, for example representing ligence, Springer-Verlag.
object-oriented database structures, default properties [10] Levesque, H. J. Foundations of a functional
of concepts, and rule-oriented knowledge structures. approach to knowledge representation. Arti cial
Perhaps, the most important aspect of this work is Intelligence 23 (1984), 155{212.
that a single representation mechanism allows for the [11] Lifschitz, V. On open defaults. In Sympo-
treatment of a large number of features, which seems sium on computational logics, J. W. Lloyd, Ed.
encouraging if we are to bridge the gap between theo- Springer-Verlag, ESPRIT Basic Research Action
retical work on frame-based languages and implement- Series, 1990, pp. 96{113.
ed systems.
[12] Lifschitz, V. Nonmonotonic databases and
epistemic queries. In Proc. of the 12th Int. Joint
References Conf. on Arti cial Intelligence IJCAI-91 (Sidney,
1991).
[1] Baader, F., and Hollunder, B. A termi- [13] Nebel, B. Reasoning and Revision in Hybrid
nological knowledge representation system with Representation Systems. Lecture Notes in Arti -
complete inference algorithm. In Proc. of the cial Intelligence. Springer-Verlag, 1990.
Workshop on Processing Declarative Knowledge,
PDK-91 (1991), Lecture Notes in Arti cial Intel- [14] Nebel, B. Terminological reasoning is inherently
ligence, Springer-Verlag. intractable. Arti cial Intelligence 43 (1990), 235{
[2] Borgida, A., Brachman, R., McGuinness, 249.
D., and Resnick, L. CLASSIC: A structural [15] Reiter, R. On closed world data bases. In Logic
data model for objects. In ACM SIGMOD Inter- and Databases, H. Gallaire and J. Minker, Eds.
national Conf. on Management of Data (1989), Plenum, 1978, pp. 119{140.
pp. 58{67. [16] Reiter, R. On asking what a database knows. In
[3] Brachman, R. J., and Levesque, H. J. The Symposium on computational logics (1990), J. W.
tractability of subsumption in frame-based de- Lloyd, Ed., Springer-Verlag, ESPRIT Basic Re-
scription languages. In Proc. of the 4th Nat. Conf. search Action Series, pp. 96{113.
on Arti cial Intelligence AAAI-84 (1984). [17] Schild, K. Towards a theory of frames and rules.
[4] Donini, F. M., Lenzerini, M., Nardi, D., Tech. rep., FB Informatik, Technische Universitat
and Nutt, W. The complexity of concept lan- Berlin, Berlin, Germany, 1989.
guages. In Proc. of the 2nd Int. Conf. on Princi- [18] Schmidt-Schau, M., and Smolka, G. At-
ples of Knowledge Representation and Reasoning tributive concept descriptions with complements.
KR-91 (1991), J. Allen, R. Fikes, and E. Sande- Arti cial Intelligence 48, 1 (1991), 1{26.
wall, Eds., Morgan Kaufmann, pp. 151{162. [19] Ullman, J. Principles of Database and Knowl-
[5] Donini, F. M., Lenzerini, M., Nardi, D., edge Base Systems, vol. 1. Computer Science
and Nutt, W. Tractable concept languages. In Press, Potomac, Maryland, 1988.
Proc. of the 12th Int. Joint Conf. on Arti cial [20] Woods, W. A. Understanding subsumption and
Intelligence IJCAI-91 (Sidney, 1991). taxomony: A framework for progress. In Princi-
[6] Donini, F. M., Lenzerini, M., Nardi, D., ples of Semantic Networks, J. Sowa, Ed. Morgan
and Schaerf, A. From subsumption to instance Kaufmann, 1991, pp. 45{94.
checking. Tech. rep., Dipartimento di Informatica [21] Yen, J., Neches, R., and MacGregor, R.
e Sistemistica, Universita di Roma \La Sapienza", CLASP: Integrating term subsumption sstems
Forthcoming. and production systems. IEEE trans. on Knowl-
[7] Doyle, J., and Patil, R. S. Two theses edge and Data Engineering 3, 1 (1991), 25{31.
of knowledge representation: Language restric-
tions, taxonomic classi cation, and the utility of

View publication stats