Catálogo de publicaciones - libros

Semistructured data has attracted much attention in database research lately. New data models, query languages and systems have been developed. However, very little research has been done on for the new data models and query languages. As a consequence, some aspects of the meaning of expressions in the database and answers to queries are not completely understood. In this paper we propose a semantics for semistructured data based on a modal logic framework used to give semantics to feature structures. A semistructured database is seen as inducing a certain model, which is then used to evaluate queries posed to the database in a first-order language. The main advantage of this approach is to give a formal basis for analyzing some possibly conflicting views of semistructured data as a model for .

The evolution of the has encouraged the development of new information gathering techniques. In order to retrieve information, it is necessary to integrate different sources. Planning techniques have been used for this purpose in the field of information gathering. A plan for information gathering is the sequence of actions that specify what information sources should be accessed so that some characteristicts, like access efficiency, are optimised. MAP is a multiagent framework that integrates Planning Agents and Information Retrieval Agents. In MAP, planning is not only used to integrate and to select information sources, but also to solve actual planning problems with information gathered from the . For instance, in an travel assistant domain, plans represent the sequence of actions an user has to follow to perform his/her trip. But also, each step in the plan informs the WebAgents which information sources should be accessed. In this paper we describe MAP and study experimentally two information retrieval characteristics: the average number of solutions retrieved depending on the WebAgents used and the allocated time limit, and the number of problems solved (those travel assistant problems for which at least one solution was retrieved).

Information extraction helps in building advanced tools for text processing applications such as electronic publishing and information retrieval. The applications differ in their requirements on the input text, output information, and available resources for information extraction. Since existing IE technologies are application specific, extensive expert work is required to meet the needs of each new application. However, today, most users do not have this expertise and thus need a tool to easily create IE systems tailored to their needs. We introduce a framework that consists of (1) an extensible set of advanced IE technologies together with a description of the properties of their input, output, and resources, and (2) a generator that selects the relevant technologies for a specific application and integrates these into an IE system. A prototype of the framework is presented, and the generation of IE systems for two example applications is illustrated. The results are presented as guidelines for further development of the framework.

In this paper we formalize UML class diagrams in terms of a logic belonging to Description Logics, which are subsets of First-Order Logic that have been thoroughly investigated in Knowledge Representation. The logic we have devised is specifically tailored towards the high expressiveness of UML information structuring mechanisms, and allows one to formally model important properties which typically can only be specified by means of qualifiers. The logic is equipped with decidable reasoning procedures which can be profitably exploited in reasoning on UML class diagrams. This makes it possible to provide computer aided support during the application design phase in order to automatically detect relevant properties, such as inconsistencies and redundancies.

Experiencing repeated computer virus attacks over the last decade has led to a number of intriguing technological challenges. The study of human immunology may help to solve these challenges and aid in the understanding of how and why computer viruses propagate. Current techniques employ heuristic systems to ferret the viral attacks and annihilate them. This technology is more suited for addressing known viruses and has been relatively successful. However, there remains a challenge as to how to deal with the virus problems of the near future, albeit unknown viral code. Ultimately, this paper will stimulate new innovations in anti-virus technology. This paper explores current research in the area of protection from computer viruses and suggests a possible framework for virus detection in such an autoimmune system. Hence, this paper exposes the mechanics of virus detection in the human immune system. Then it suggests techniques that could be used to identify or detect unknown viruses by distinguishing between “self” and foreigner.

Layout analysis is the process of extracting a hierarchical structure describing the layout of a page. In the document processing system WISDOM++ the layout analysis is performed in two steps: firstly, the global analysis determines possible areas containing paragraphs, sections, columns, figures and tables, and secondly, the local analysis groups together blocks that possibly fall within the same area. The result of the local analysis process strongly depends on the quality of the results of the first step. In this paper we investigate the possibility of supporting the user during the correction of the results of the global analysis. This is done by allowing the user to correct the results of the global analysis and then by learning rules for layout correction from the sequence of user actions. Experimental results on a set of multi-page documents are reported.

In this paper, we propose to provide KDD end-users with a tool which can not only discover the necessary knowledge for supporting their decision making processes but also present it in an immediately and intuitively understandable way. The presentation method is based on natural language generation (NLG) techniques. It has been implemented into the agent DM-PA which, starting from an XML representation of the discovered knowledge, can generate the description of the application of this knowledge to a specific decision context by using different presentation modalities. The presentation agent has been interfaced with INGENS, a prototypical GIS with capabilities of knowledge discovery. Results of the first integration prototype are discussed.

Knowledge engineering in AI planning is the process that deals with the acquisition, validation and maintenance of planning domain models, and the selection and optimisation of appropriate planning machinery to work on them. Our aim is to research and develop rigorous methods for the acquisition, maintenance and validation of planning domain models. We aim to provide a tools environment suitable for use by domain experts in addition to experts in the field of AI planning. In this paper we describe such a method and illustrate it with screen-shots taken from an implemented Graphical Interface for Planning with Objects system called GIPO. The GIPO tools environment has been built to support an object centred approach to planning domain modelling. The principal innovation we present in this paper is a process of specifying domain operators that abstracts away much of the technical detail traditionally required in their specification. Such innovations we believe could ultimately open up the possibility of bringing planning technology to a wider public.

Though extensions of the relational model of data have been proposed to handle probabilistic information, there has been no work to date on handling aggregate operators in such databases. In this paper, we show how classical aggregation operators (like COUNT, SUM, etc.) as well as other statistical operators (like weighted average, variance, etc.) can be defined as well as implemented over probabilistic databases. We define these operations, develop a formal linear program model for computing answers to such queries, and then develop a generic algorithm to compute aggregates.

From statistics, sampling technics were proposed and some of them were proved to be very useful in many database applications. Rather surprisingly, it seems these works never consider the preservation of data semantics. Since functional dependencies (FDs) are known to convey most of data semantics, an interesting issue would be to construct samples preserving FDs satisfied in existing relations.

To cope with this issue, we propose in this paper to define Informative Armstrong Relations (IARs); a relation is an IAR for a relation r if is a subset of and if FDs satisfied in are exactly the same as FDs satisfied in . Such a relation always exists since is obviously an IAR for itself; moreover we shall point out that small IARs with interesting bounded sizes exist. Experiments on relations available in the KDD archive were conducted and highlight the interest of IARs to sample existing relations.