CENTRIA research areas

Knowledge Representation and Reasoning, and Logic Programming (KRRLP)

This area has been focusing its activities on the following main topics: foundational research in the area of rational computational logic agents, logic programs and knowledge base updates; a general framework for integrating several reasoning forms (including fuzzy-logic, possibilistic logic, probabilistic systems, and non-monotonic logics); distributed tabling and revision systems; computational models and their implementation for a parallel and distributed logic programming language.

The current aim of this area is to further develop the work on Logic Programming (LP) and its application to Knowledge Representation and Reasoning (KRR) that had been pursued in the past in CENTRIA. The work focus on the topics:

foundational research in extensions of LP for KRR, including extensions to deal with updates and dynamic knowledge, and extensions to deal with inconsistent and uncertain (probabilistic and fuzzy) knowledge and attending procedures;
applications to LP based knowledge representation, mainly to the areas of rational computational logic agents, legal knowledge representation and reasoning, and intelligent information systems;
implementations of LP systems.

These topics have strong relations amongst themselves, in that implementations are guided by the foundational results and will be used in the applications. They also relate to the work of other projects of CENTRIA, mainly to those of Semantic Engines for the Web and of Intelligent Information Systems, via the applications topic.

In the near future we intend to:

develop a unified logic programming language for specifying updates in knowledge bases. The language should have enough expressivity for declaratively specifying the knowledge about a given domain, as well as the knowledge of how it evolves, including self-evolution. Besides the establishment of the language and corresponding semantics, it is also our goal to develop proof procedures for reasoning with this language, to implement them, and to study its application to various domains, including legal reasoning and rational epistemic agents.
integrate in the LP language several uncertainty reasoning forms, including fuzzy-logic, possibilistic logic, probabilistic systems, Bayesian networks, and non-monotonic and paraconsistent logics. Again, this work will span from the foundations and declarative semantics, to the definition of proof systems and corresponding implementations.
as a basis for the two goals above, there is the need for efficient implementations of logic programming. It will also be a goal of this area to promote further developments of the GNU Prolog system for making it more efficient and amenable for the other goals. Among these developments are: distributed tabling; ISO modules; contextual LP, multiple threads.

Intelligent Information Systems (IIS)

In this area, research work covers the following topics: semantic web definition, tools for semantic web based integration of heterogeneous databases, intelligent agents for automatic classification of documents, definition of semantic web ontologies, and natural language dialogue systems for information retrieval from intelligent Information systems. Somewhat related to Intelligent Information Systems, is the work done in the area of text mining. However, we opted for its inclusion in the subsection of Soft Computing and Constraints.

Here, we plan to develop the topics: Semantic Web definition, Tools for semantic web based integration of heterogeneous databases, intelligent agents for automatic classification of documents and definition of semantic web ontologies, natural language dialogue systems for information retrieval from intelligent Information systems and Data Warehouse Design and Query.

We intend to develop tools and applications in the following areas: development and integration of heterogeneous Databases in Information Systems; Intelligent agents for automatic classification of documents and definition of semantic web ontologies; Natural language dialogue systems for information retrieval from intelligent Information systems; and Data Warehouse Design and Query.

The development and integration of heterogeneous Databases in Information Systems will be done mainly through use of ISCO extended with some new features enabling explicit tense representation. Results here will be used in other project areas.

We aim to build Intelligent agents for automatic classification of documents and for automatic definition of semantic web ontologies. This will enable us to develop applications in the law domain allowing users in general, and Portuguese citizens in particular, to have an easier access to legal information. This tool will enable us to build a dialogue system for controlling the user interaction with an Information Retrieval system. We expect that the same dialogue system can be used as an interface to the Information System of Univ. Évora enabling the users an easy access to the information.

Some tools applications in the area of Data Warehouse Design and Query will also be done.

In Informations systems we expect projects to grow into several new areas, including: meeting and class scheduling, ERP applied to higher education institutions, natural language querying, integration into an UML framework and use of UML editors.

In Information Retrieval we intend to build specialized tools for processing Portuguese documents: solving anaphoras, automatic classification, and clustering. And to develop a cooperative multimodal information-retrieval system which could be applied to several domains of knowledge. In particular, it could be a pplied to other areas where there is a need for intelligent document management.

In natural language dialogues system we expect to build a module for automatically building the knowledge base for the semantic/pragmatic sentences interpretation from the information system description and the Semantic Web ontology; a module for the semantic interpretation via the abduction of predicates from the ISCO language and through the interrogation of the databases.

In Designing Data Warehouses and Query we expect to build the following prototype tools: Computer Assisted Multidimensional Modeling CAMM; Computer Assisted Physical Design for Multidimensional Modeling (including the design of aggregated tables); an XML based for Scheduling Problems Web Services and one View Maintenance of ORDB.

Soft Computing and Constraints (SCC)

The planned research activity will not only extend ongoing work, but also explore new directions in both fundamental and applied research. In the latter, and enlarging the scope of our long term interest in applications of AI in Medicine, we intend to address more applications in Bioinformatics, some of which are already under way. This is a scientific area of great importance at present, yielding rich sets of data that are a challenge to both CP and ML.

Recent work in SCC include: integration of local search and constraint propagation, improvement of interaction of constraint propagation techniques with Computational Geometry methods, development of sets constraints solver, approaching different optimisation problems, work on global constraints, spatial constraints, and over-constrained problems, development of search techniques...

CENTRIA team covers a sufficiently large spectra of related research topics in the areas of constraint satisfaction and optimisation, as well as automated learning and data mining, so as to make it possible to share experiences and take advantage of the synergies and cross-fertilisation that is possible within this area, as well as with other areas in CENTRIA.

Current research topics include:

Modeling biophysical Systems with non-linear constraints over continuous domains
Constraint Solving over Finite Domains and Boolean clauses
Architecture for Distributed Constraint Solvers
Concept Learning and Datamining

We intend to extend previous research on multi-valued logics for digital circuits to applications of logic-based agents. In this topic of agents, we intend to study the integration of Fuzzy Constraint Satisfaction Problems (FCSP) and Multi-Agent Systems (MAS) in such a way that problem solving in MAS is the result of an interaction between agents solving a FCSP. Issues to be addressed include distributed consistency reinforcement: (constraint propagation among agents) and the global degree of satisfaction will be an aggregation of the agents' degree of satisfaction or a global measure, obtained by other means external to the system. Other research work will be continued regarding the learning of grammatical constructions via sensorimotor representations. All these topics should provide fruitful interaction with the KRRLP area.

The planned activity will extend ongoing work on both fundamental and applied research in the areas of Constraint programming and Machine Learning. In regard to the former, we will address new architectures for finite domain constraint solvers and new methods for constraint solving involving local search, fuzzy constraints, and constraints over continuous domains. Fundamental research in machine learning will focus on new models for concept learning via fuzzy clustering, new algorithms for self-organising maps data mining and inductive logic programming as well as neural networks coupled with genetic algorithms.

In regard to the latter, and thereby enlarging the scope of our long term interest in applications of AI in Medicine, we intend to address a number of applications in Bioinformatics, which will bring together research results in both constraint programming and machine learning. A number of new applications are sought to explore machine learning techniques, namely in the interpretation of oceanographic data, in the management and intelligent access to music datawarehouses, and in more general search texts and datamining in web pages.

Departamento de Informática, FCT/UNL
Quinta da Torre 2829-516 CAPARICA - Portugal
Tel. (+351) 21 294 8536 FAX (+351) 21 294 8541