Catálogo de publicaciones - libros

This paper describes a theoretical framework for inducing knowledge from incomplete data sets. The general framework can be used with any formalism based on a lattice structure. It is illustrated within two formalisms: the attribute-value formalism and Sowa’s conceptual graphs. The induction engine is based on a non-supervised algorithm called default clustering which uses the concept of stereotype and the new notion of default subsumption, inspired by the default logic theory. A validation using artificial data sets and an application concerning the extraction of stereotypes from newspaper articles are given at the end of the paper.

A viable ontology engineering methodology requires supporting domain experts in gradually building and managing increasingly complex versions of ontological elements and their converging and diverging interrelationships. Contexts are necessary to formalise and reason about such a dynamic wealth of knowledge. However, context dependencies introduce many complexities. In this article, we introduce a formal framework for supporting context dependency management processes, based on the DOGMA framework and methodology for scalable ontology engineering. Key notions are a set of context dependency operators, which can be combined to manage complex context dependencies like articulation, application, specialisation, and revision dependencies. In turn, these dependencies can be used in context-driven ontology engineering processes tailored to the specific requirements of collaborative communities. This is illustrated by a real-world case of interorganisational competency ontology engineering.

Classifications have been used for centuries with the goal of cataloguing and searching large sets of objects. In the early days it was mainly books; lately it has also become Web pages, pictures and any kind of digital resources. Classifications describe their contents using natural language labels, an approach which has proved very effective in manual classification. However natural language labels show their limitations when one tries to automate the process, as they make it very hard to reason about classifications and their contents. In this paper we introduce the novel notion of , as a graph structure where labels are written in a propositional concept language. Formal Classifications turn out to be some form of lightweight ontologies. This, in turn, allows us to reason about them, to associate to each node a normal form formula which univocally describes its contents, and to reduce document classification and query answering to reasoning about subsumption.

The goal of is the development of new technologies and mechanisms to support the integration, evolution and matching of data models at the conceptual and logical design level. Such tasks are to be performed by means of a set of model management which work on models and their elements, without being restricted to a particular metamodel (e.g. the relational or UML metamodel).

We propose that generic model management should employ a generic metamodel (GMM) which serves as an abstraction of particular metamodels and preserves as much of the original features of modeling constructs as possible. A naive generalization of the elements of concrete metamodels in generic metaclasses would lose some of the specific features of the metamodels, or yield a prohibitive number of metaclasses in the GMM. To avoid these problems, we propose the in which each model element is with a set of role objects that represent specific properties of the model element. Roles may be added to or removed from elements at any time, which enables a very flexible and dynamic yet accurate definition of models.

Roles expose to operators different views on the same model element. Thus, operators concentrate on features which affect their functionality but may remain agnostic about other features. Consequently, these operators can use polymorphism and have to be implemented only once using , and not for each specific metamodel. We verified our results by implementing and a selection of model management operators using our metadata system ConceptBase.

A data warehouse (DW) is a database that integrates data from external data sources (EDSs) for the purpose of advanced analysis. EDSs are production systems that often change not only their contents but also their structures. The evolution of EDSs has to be reflected in a DW that integrates the sources. Traditional DW systems offer a limited support for the evolution of their structures. Our solution to this problem is based on a multiversion data warehouse (MVDW). Such a DW is composed of the sequence of persistent versions, each of which describes a schema and data within a given time period. The management of the MVDW requires a metadata model that is much more complex than in traditional data warehouses. In our approach and prototype MVDW system, the metadata model contains data structures that support: (1) monitoring EDSs with respect to content and structural changes, (2) automatic generation of processes monitoring EDSs, (3) applying discovered EDS changes to a selected DW version, (4) describing the structure of every DW version, (5) querying multiple DW versions at the same time and presenting the results coming from multiple versions.

The Semantic Web envisions a world-wide distributed architecture where computational resources will easily inter-operate to coordinate complex tasks such as query answering. Semantic marking up of web resources using ontologies is expected to provide the necessary glue for making this vision work. Using ontology languages, (communities of) users will build their own ontologies in order to describe their own data. Adding semantic mappings between those ontologies, in order to semantically relate the data to share, gives rise to the Semantic Web: data on the web that are annotated by ontologies networked together by mappings. In this vision, the Semantic Web is a huge semantic peer data management system. In this paper, we describe the peer data management systems that promote a ”simple is beautiful” vision of the Semantic Web based on data annotated by RDFS ontologies.

Ontologies are an important technology for the Semantic Web. In different areas ontologies have already been developed and many of these ontologies contain overlapping information. Often we would therefore want to be able to use multiple ontologies. To obtain good results, we need to find the relationships between terms in the different ontologies, i.e. we need to align them. Currently, there exist a number of systems that support users in aligning ontologies, but not many comparative evaluations have been performed and there exists little support to perform such evaluations. However, the study of the properties, the evaluation and comparison of the alignment strategies and their combinations, would give us valuable insight in how the strategies could be used in the best way. In this paper we propose the KitAMO framework for comparative evaluation of ontology alignment strategies and their combinations and present our current implementation. We evaluate the implementation with respect to performance. We also illustrate how the system can be used to evaluate and compare alignment strategies and their combinations in terms of performance and quality of the proposed alignments. Further, we show how the results can be analyzed to obtain deeper insights into the properties of the strategies.

Integrating data mining techniques into database systems has gained popularity and its significance is well recognized. However, the performance of SQL based data mining is known to fall behind specialized implementations. Reasons for this are among others the prohibitive nature of the cost associated with extracting knowledge as well as the lack of suitable declarative query language support. Recent studies have found that for association rule mining and sequential pattern mining with carefully tuned SQL formulations it is possible to achieve performance comparable to systems that cache the data in files outside the DBMS. However, most of the previous pattern mining methods follow the method of , which still encounters problems when a sequential database is large and/or when sequential patterns to be mined are numerous and long.

In this paper, we present a novel SQL based approach that we recently proposed, called (PROjection Sequential PAttern Discovery). fundamentally differs from an Apriori-like candidate set generation-and-test approach. This approach is a pattern growth-based approach without candidate generation. It grows longer patterns from shorter ones by successively projecting the sequential table into subsequential tables. Since a projected table for a sequential pattern contains all and only necessary information for mining the sequential patterns that can grow from , the size of the projected table usually reduces quickly as mining proceeds to longer patterns. Moreover, a depth first approach is used to facilitate the projecting process in order to avoid creating and dropping costs of temporary tables.