Catálogo de publicaciones - libros

The paper presents INFOMIX a successful application of ASP technology to the domain of Data Integration. INFOMIX is a novel system which supports powerful information integration, utilizing the ASP system DLV. While INFOMIX is based on solid theoretical foundations, it is a user-friendly system, endowed with graphical user interfaces for the average database user and administrator, respectively. The main features of the INFOMIX system are: (i) a comprehensive information model, through which the knowledge about the integration domain can be declaratively specified, (ii) capability of dealing with data that may result incomplete and/or inconsistent with respect to global constraints, (iii) advanced information integration algorithms, which reduce (in a sound and complete way) query answering to cautious reasoning on disjunctive Datalog programs, (iv) sophisticated optimization techniques guaranteeing the effectiveness of query evaluation in INFOMIX, (v) a rich data acquisition and transformation framework for accessing heterogeneous data in many formats including relational, XML, and HTML data.

We associate optimality theory with abduction and preference handling. We present linguistic problems that appear in the study of dialects as new application of abduction and preference handling. Differences in German dialects originate from different rankings of linguistic constraints which determine the well-formedness of expressions within a language. We introduce a framework for analyzing differences in German dialects by abduction of preferences. More precisely, we will take the perspective of a linguist and reconstruct dialectal variation as abduction problem: Given an observation that a sentence is found as grammatically correct, abduct the underlying constraint ranking. For this, we give a new definition for the determination of optimal candidates for total orders with indifferences. Additionally, we give an encoding for the diagnosis front-end of the system.

We describe an application of Abductive Logic Programming (ALP) to the analysis of an important class of DNA microarray experiments. We develop an ALP theory that provides a simple and general model of how gene interactions can cause changes in observable expression levels of genes. Input to the procedure are the observed microarray results; output are hypotheses about possible gene interactions that explain the observed effects. We apply and evaluate our approach on microarray experiments on and .

The integration of preferences into Answer Set Programming (ASP) constitutes an important practical device for distinguishing certain preferred answer sets from non-preferred ones. Up to now, the preference semantics we are considering in this system description were incorporated into answer set solvers either by meta-interpretation [3] or by pre-compilation front-ends [2]; therefore, such kinds of preferences were never integrated into the core existing ASP solvers.

A number of answer set solvers have been proposed in recent years, such as , DLV, Cmodels, and ASSAT. Most existing ASP solvers have been extended to provide front-ends that are suitable to encode specialized forms of knowledge-e.g., weight-constraints, restricted forms of optimization, front-ends for planning and diagnosis. These features allow declarative solutions in speci.c application domains. However, this is not completely satisfactory:

– The development of an ASP program is viewed as a “monolithic” process. Most ASP systems offer only a batch approach to execution of programs-programs are completely developed, “compiled”, executed, and finally answer sets are proposed to the user. The process lacks any levels of interaction with the user. In particular, it does not directly support an interactive development of programs (as in Prolog), where one can immediately explore the results of simply adding/removing rules.

– ASP programmers can control the computation of answer sets through the rules that they include in the logic program. Nevertheless, ASP systems offer very limited capabilities for reasoning on the of answer sets associated to a program-e.g., to perform selection of models according to user-defined criteria or to compare models. These activities are important in many application domains-e.g., to express soft constraints, to support preferences when using ASP to perform planning.

– ASP solvers are independent systems; interaction with other languages can be performed only through complex, low level APIs; this prevents programmers from writing programs that manipulate ASP programs and answer sets as first-class citizens. E.g., we wish to write programs in a high-level language (Prolog in this case), which are capable to access ASP programs, modify their structure (by adding or removing rules), and access and reason with answer sets. This type of features is essential in many domains-e.g., automatically modify the plan length in a planning problem.

The of disjunctive logic programs (DLPs) is based on minimal models [5,12] which makes atoms appearing in a disjunctive program false by default. This is often desirable from the knowledge representation point of view, but certain domains become awkward to formalize if all atoms are blindly subject to minimization. In contrast to this, [11] provides a re.ned notion of minimal models as it distinguishes and atoms in addition to those being falsified. This eases the task of knowledge presentation in many cases. For example, it is straightforward to formalize Reiter-style [13] for digital circuits using parallel circumscription.

We describe a new software, , that uses pseudo-boolean constraint solvers ( solvers) to compute stable models of logic programs with weight atoms. To this end, converts ground logic programs to propositional theories with weight atoms so that stable models correspond to models. Our approach is similar to that used by and . However, unlike these two systems, does not compile the weight atoms away. Preliminary experimental results on the performance of are promising.

We present a monitoring tool for plan execution in non-deterministic environments, which are described in an action language, based on non-monotonic logic programming. Thanks to it, deviations of concrete executions from expected ones can be detected, and diagnostic explanations in terms of unsuccessful action executions can be obtained. The latter may be exploited for execution recovery, and may help in rectifying an incoherent view of the planning domain.

We present a new answer set solver ++. Distinguishing features include its treatment of heads and bodies equitably as computational objects and a new hybrid lookahead. ++ is close to being competitive with state-of-the-art answer set solvers, as demonstrated by selected experimental results.

In [2], we presented a system called - for computing answer sets of logic programs with aggregates. The implementation of - relies on the use of an external constraint solver (ECLiPSe) to deal with aggregate literals and requires some modifications to the answer set solver used in the experiment (). In general, the system is capable of computing answer sets of arbitrary programs with aggregates, i.e., there is no syntactical restrictions imposed on the inputs to the system. This makes - different from (built BEN/5/23/04) [1], which deals with stratified programs only. -, however, is based on a semantics that does not guarantee minimality of answer sets. Furthermore, our experiments with - indicate that the cost of communication between the constraint solver and the answer set solver proves to be significant in large instances.