EPIA 91: 5th Portuguese Conference on Artificial Intelligence Albufeira, Portugal, October 1–3, 1991 Proceedings

301 Pages·1991·4.935 MB·English

by  Miguel Filgueiras, Ana Paula Tomás (auth.), Pedro Barahona, Luís Moniz Pereira, António Porto (eds.)

Lecture Notes in Artificial Intelligence 541
Subseries of Lecture Notes in Computer Science
Edited by J. Siekmann
Lecture Notes in Computer Science
Edited by G. Goos and J. Hartmanis
E Barahona L. Moniz Pereira A. Porto (Eds.)
EPIA 91
5th Portuguese Conference on Artificial Intelligence
Albufeira, Portugal, October 1-3, 1991
Proceedings
galreV-regnirpS
Berlin Heidelberg NewYork
London Paris Tokyo
Kong Hong Barcelona
Budapest
Series Editor
J6rg Siekmann
Institut ftir Informatik, Universit~t Kaiserslautern
Postfach 3049, W-6750 Kaiserslautern, FRG
Volume Editors
Pedro Barahona
Luis Moniz Pereira
Ant6nio Porto
Departamento de Inform~itica, Universidade Nova de Lisboa
2825 Monte da Caparica, Portugal
CR Subject Classification (1991): 1.2.3-4, 1.2.6, 1.2.8
ISBN 3-540-54535-2 Springer-Verlag Berlin Heidelberg New York
ISBN 0-387-54535-2 Springer-Verlag New York Berlin Heidelberg
This work is subject to copyright. All rights are reserved, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, re-use of
illustrations, recitation, broadcasting, reproduction on microfilms or in otherw aysa,n d
storage in data banks. Duplication of this publication or parts thereof is only permitted
under the provisions of the German Copyright Law of September 9, 1965, ini ts current
version, and a copyright fee must always be paid. Violations fall under the prosecution
act of the German Copyright Law.
© Springer-Verlag Berlin Heidelberg 1991
Printed in Germany
Typesetting: Camera ready by author
Printing and binding: Druckhaus Beltz, Hemsbach/Bergstr.
2145/3140-543210 - Printed on acid-free paper
Preface
The Portuguese Association for Artificial Intelligence (Associa~o Portuguesa
Para a Intelig~ncia Artificial-APPIA) has been organizing Portuguese
Conferences on Artificial Intelligence, now held every second year, since 1985.
The present Fifth Conference on Artificial Intelligence, like the previous one, was
aimed at an international audience and at a high standard regarding accepted
communications.
The international status of the Conference can be assessed from the fact that 62
contributions from 31 countries were received (26 contributions from Portugal).
All of them were reviewed by at least three researchers of the appropriate
scientific areas, and the reviews closely scrutinized by the Program Committee.
To guarantee a high scientific standard, the Program Committee decided to
accept only those papers significantly better than merely acceptable, and in fact
most of them were judged definitely acceptable. As a result, only 20 papers were
accepted and included in these proceedings.
We would like to thank all the people who made important contributions to the
technical quality of this Conference, namely the Program Committee members
and the referees, listed elsewhere in these proceedings.
We also thank our invited speakers Hassan Ait-Kaci, Robin Cooper, Mark
Drummond, Catherine Lassez Vladimir Lipschitz, and Werner Schimanovich
for their presentations, which significantly increased the interest and quality of
the Conference.
Finally, we would like to thank the institutions who have contributed (financially
or otherwise) to the organization of this Conference, namely Funda~o Calouste
Gulbenkian, Instituto Nacional de Investiga~ao Cientifica, Junta Nacional de
Investiga~o Cientifica e Tecnol6gica, Funda~o Luso-Americana para o
Desenvolvimento, Banco Espirito Santo e Comercial de Lisboa, Caixa Geral de
Dep6sitos, Companhia Portuguesa Radio Marconi, Alcatel Portugal, IBM
Portuguesa, Centro de Intelig~ncia Artificial do UNINOVA, and Departamento
de Inform~tica da Universidade Nova de Lisboa.
Lisbon, July 1991 Luis Moniz Pereira
Ant6nio Porto
Pedro Barahona
IV
Conference Chair Program Chair
Pedro Barahona Lufs Moniz Pereira
Ant6nio Porto
Program Committee
Pedro Barahona Universidade Nova de Lisboa
Pavel Brazdil Universidade do Porto
Helder Coelho Universidade T6cnica de Lisboa
Gabriel Pereira Lopes Universidade Nova de Lisboa
Jo~o Pav~o Martins Universidade T6cnica de Lisboa
Fernando Pereira ATT Bell Laboratories, New Jersey
Luis Moniz Pereira Universidade Nova de Lisboa
Ant6nio Porto Universidade Nova de Lisboa
Referees
.J Alegria .J .J Alferes P. Barahona
A. Borges deA lmeida I. Braha P. Brazdil
R. Camacho L. Camarinha de Matos .J Cam6es Silva
A. Cardoso .J Cohen H. Coelho
R. Cooper E. Costa .J Cunha
L. Damas .J Duarte do Amaral .J Esgalhado Valenqa
.J Falc~o e Cunha M. Filgueiras G. Gaspar
M. Gomes P. Guedes de Oliveira T. Hichey
.J Hirschberg .J M. Jacquet D. Johnson
.J P. Leal .J Legatheaux Martins M. Mamede
A. Matos M. Matos F. Menezes
L. Moniz Pereira L. Monteiro E. Morgado
.J Nunes Aparicio A. Odlyzko E. Oliveira
.J N. Oliveira J. Pav~o Martins F. Pereira
G. Pereira Lopes I. Pimenta Rodrigues A. Pinho
A. Porto M. Pr6spero dos Santos C. Queiroz
C. Ramos R. Ribeiro A. Sernadas
C. Sernadas L. Torgo A. Tra~a de Almeida
Table of Contents
Constraints
Solving Linear Constraints on Finite Domains Through Parsing .......................................
M. Filgueiras, A. P. Tom,is
Centro de Inform~itica, Universidade do Porto, Portugal
Constraint Solving in Finite Domains under User Control .................................................. 17
.F Menezes, P. Barahona
Departamento de InformStica, Universidade Nova de Lisboa, Portugal
A New Method for Solving Linear Constraints on the Natural Numbers ........................ 30
A. P. Tom,is, M. Filgueiras
Centro de Inform~tica, Universidade do Porto, Portugal
A Constraint-Based Fuzzy Inference System ............................................................................. 45
K. Lano
Computing Laboratory, Oxford University, United Kingdom
A Constraint-Based Language for Querying Taxonomic Systems ........................................ 60
M. Mamede, L. Monteiro
Departamento de Informatica, Universidade Nova de Lisboa, Portugal
Search
Heuristic Parsing and Search Space Pruning ............................................................................. 76
.J A. Mendes, P. R. Henriques
Universidade do Minho, Braga, Portugal
Wave-shaping in Multiprocessor Bidirectional Heuristic State Space Search .................. 92
P. C. Nelson, A. A. Toptsis
Department of Electrical Engineering and Computer Science, University of Illinois, USA;
Department of Computer Science and Mathematics, York University, Canada
Knowledge Representation
The Extended Stable Models of Contradiction Removal Semantics .................................... 501
L. Moniz Pereira, .J .J Alferes, .J N. Aparicio
Departamento de Informatica, Universidade Nova de Lisboa, Portugal
Modelling a Rational Cognitive Agent in SNePS .................................................................... 021
D. Kumar, .S C. Shapiro
Department of Computer Science, SUNY at Buffalo, USA
Semantics of Property Inheritance in a Hierarchic System with Explicit Negation ......... 531
G. David, A. Porto
Departamento de Inform~tica, Universidade Nova de Lisboa, Portugal
iiiv
Temporal Reasoning
Time In Confluences: Dealing with Delays for Consistency-Checking ............................... 151
A. Cardoso, E. Costa
Laborat6rio de Inform~itica e Sistemas, Universidade de Coimbra, Portugal
A Temporal Representation for Imperatively Structured Plans of Actions ...................... 561
E. Rutten
CWI, Amsterdam, The Netherlands
Maximal Intervals: An Approach to Temporal Reasoning ................................................... 081
C. Ribeiro, A. Porto
Departamento de Inform~itica, Universidade Nova de Lisboa, Portugal
Planning
Consistency Driven Planning ....................................................................................................... 591
M. Decker, G. Moerkotte, H. M~iller, .J Posegga
IPD - Fakult/it f/Jr Informatik, Universit/it Karlsruhe, Germany
An Efficient Approach to PlannIng in Assembly Tasks .......................................................... 210
C. Ramos, E. Oliveira
Faculdade de Engenharia, Universidade do Porto, Portugal
Diagnosis and Repair
Towards a Theory of the Repair Process ..................................................................................... 222
G. Friedrich, G. Gotflob, W. Nejdl
Institut f/Jr Informationssysteme, Technische Universit/it Wien, Austria
Declarative Source Debugging ...................................................................................................... 237
M. Calejo, L. Moniz Pereira
Departamento de Inform~itica, Universidade Nova de Lisboa, Portugal
Learning
A Neural Approach to Data Compression and Classification ............................................... 250
K. P. Kratzer
Fachhochschule Ulm, Germany
Generalization for a Propositional Calculus: A Constraints-Based Approach .................. 264
R. Vorc'h
IRISA, Rennes, France
Adaptive Learning Using a Qualitative Feedback Loop ......................................................... 278
L. Winkelbauer, C. Stary
International Institute for Applied Systems Analysis, Laxenburg, Austria;
School of Computer Science, Florida International University, USA
Solving Constraints on Linear Finite Domains Through
*gnisraP
Miguel Filgueiras, Ana Paula Tom£s
Centro de Inform~tica, Universidade do Porto
R. do Campo Alegre 823, 4100 Porto, Portugal
email: [email protected], [email protected]
Abstract
In this paper we present results from ongoing research which allows the
use of parsing methods to solve a particular kind of constraints, namely linear
constraints on finite domains. Solving this kind of constraints is equivalent
to solving systems of linear Diophantine equations on a finite subset of the
naturals. We associate, to such a system, a definite-clause grammar that
can be used to enumerate its solutions, and define a class of grammars, the
connected grammars, for which the set of successful derivations covers the set
of non-negative solutions of the associated system. This definition is based
on a study of cycles in context-free grammars using compiler construction
concepts and techniques.
1 Introduction
The emergence of Constraint Logic Programming languages, namely Prolog III
[Colmerauer 1987] and CHIP [Dincbas et al. 1988], [van Hentenryck 1989], and
the definition of the CLP-scheme [Jaffar et al. 1986], [Jaff~r and Lassez 1987],
increased and renewed the importance of the study of methods for handling con-
straints and, in particular, methods for solving constraints [Cohen 1990].
*The research described in this paper saw partially supported by Instituto NacionM de In-
Cientifica. vestiga~go
In this paper we present results from ongoing research which allow the use
of parsing methods to solve a particular kind of constraints, namely linear con-
straints on finite domains 1 . Solving this kind of constraints is equivalent to solving
systems of linear Diophantine equations on a finite subset of the naturals. It is
easily seen that such a system may be always transformed into an equivalent one
with positive coefficients by an appropriate use of the bounds for the unknowns.
We show that there is a natural correspondence between a system and a definite-
clause grammar (DCG) Pereira and Warren 1980, what enables us to trivially
map the problem of solving linear constraints on finite domains into a parsing
problem, and to try to use efficient parsing algorithms to solve the first one. How
interesting this may turn out in practice depends on the results of a compari-
son with other methods. The soundness requirement that the set of successful
grammar derivations covers the solution set of the system is proved for the class
of connected grammars, a sub-class of context-free grammars introduced below.
The results presented can be easily extended to a subset of the DCGs.
The next section gives a formal presentation of how successful derivations
in a grammar can be related with a system of non-negative linear Diophantine
(NNLD, for short) equations. In Section 3 notions concerning cycles in grammars
are introduced that allow the definition, in Section 4 of the class of connected
grammars. In order to prove (Section 6) that for this class the set of successful
derivations covers the set of solutions of the associated system of equations, some
other notions and results concerning subtrees and multisets of subtrees (forests)
are described in Section .5 Finally, and before the Conclusions, an overview of
how parsing can be used to solve linear constraints on finite domains is given.
2 Derivations and Systems of Equations
To every context-free grammar G it is possible to associate a system of linear
equations on non-negative integers in which each variable stands for the number
of applications of each production rule of G in a successful derivation. Let G =
(Vt, Vn,P,S) and for every iv E P, 1 < i < p let nl be the number of times ri
is applied during a derivation of a string in L(G), the language generated by G.
Let
• ip(A) denote the set of indices i of the rules rl whose left-hand side is A ,2
eeS1 also sarieugliF ,0991 sarieugliF and Tom~s .0991
~Without loss of generality it will be assumed that rof every non-terminal A E If,, ip(A) ~ .~
• rhs(r) be the right-hand side of rule r,
• occ(X, (r) be the number of occurrences of symbol X in the string ,ro and
• n(x) be the number of occurrences of the terminal z in the string to be
parsed.
In a successful derivation of a string w E L(G), the number of times a non-
terminal symbol is expanded must be equal to the number of times it is produced
(the start symbol obviously has one more occurrence). Similarly, the number of
times a terminal symbol occurs in w must be equal to the number of times it is
produced. As a consequenceo f these facts given a string w E L(G) the following
equations must hold:
non-terminal equations:
•
for S, the start symbol of G
P
ni - ~_, occ(S, rhs(rk)) * nk = 1
ieip(S) k=l
for every non-terminal A(¢ S) q nV
P
ni - ~ occ(A, rhs(rk)) . nk = 0
iEip(A) k=l
• terminal equations:
for every terminal a E tV
p
occ(a, rhs( k)) • = oct(.,
k=l
A particular string determines a set of solutions to the system. However, a
solution in this set may either correspond to a successful derivation of a different
string or to no derivation at all. The first situation arises because in the system
there is no information about the order of symbols. The second one originates
from the existence of cycles in the grammar: a solution may correspond to a
sucessful derivation plus a certain number of cycles.
When our goal is to solve a system of NNLD equations by parsing, the first
case is not really problematic because we can fix the order and the string to be
parsed. In contrast, the second means that the set of solutions is not covered by