Vedecké časopisy a ročenky vydávané na pôde SAV

Dynamic aspects of knowledge representation has been tackled recently by a variety of approaches in the logic programming style. We consider the approaches characterized by the causal rejection principle (if there is a conflict between rules, then more preferred rules override those less preferred). A classification and a comparison of the approaches is presented in the paper. We compare them also to our own approach based on Kripke structures.

Temporal ontology for representing uncertainly specified time periods is presented. Mature approaches like Allen interval relations are combined with introduction of time granularity and time uncertainty concepts. The ontology is applicable both as a static data representation and for logical data inference. Logical conclusions can be derived using an automated inference system. Uncertainty parametrization was developed for handling the domain specific uncertainty characteristics. Temporal statements containing the most frequent expressions in the domain of cultural heritage preservation are identified and categorized with respect to their accuracy. A temporal inference system is implemented using OCML language. Consistency checks can find non-causal data clusters and lead to improving current event data. Finally, resource annotation with Dynamic Narrative Authoring Tool utilizing temporal inference is presented.

Composition of simpler web services into custom applications is understood as promising technique for information requests in a heterogeneous and changing environment. This is also relevant for applications characterised as deductive web mining (DWM). We suggest to use problem-solving methods (PSMs) as templates for composed services. We developed a multi-dimensional, ontology-based framework, and a collection of PSMs, which enable to characterise DWM applications at an abstract level; we describe several existing applications in this framework. We show that the heterogeneity and unboundedness of the web demands for some modifications of the PSM paradigm used in the context of traditional artificial intelligence. Finally, as simple proof of concept, we simulate automated DWM service composition on a small collection of services, PSM-based templates, data objects and ontological knowledge, all implemented in Prolog.

In this paper we deal with personalized navigation in an open information space. Our aim is to support effective orientation in increasing amount of information accessible through the Web. We present a method for personalized navigation based on social navigation where the information space is represented by an ontology. Navigational information is obtained by following user footsteps. It is attached to information fragment mapped to the user goal and to description of this goal using an ontology. This information is used later to show the way to similar goals. We use ontology representation of the information space that supports the effective navigation and the navigational ability to deal with frequent changes of information content in open environments. We demonstrate the proposed method in the context of developed software tool PENA for personalized navigation support in labor supply domain.

The paper focuses on the field of automatic extraction of information from texts and text document categorisation including pre-processing of text documents, which can be found on the Internet. In the frame of the presented work, we have devoted our attention to the following issues related to text categorisation: increasing the precision of categorisation algorithm results with the aid of a boosting method; searching a minimum number of decision trees, which enables the improvement of the categorisation; the influence of unlabeled data with predicted categories on categorisation precision; shortening click streams needed to access a given web document; and generation of key words related with a web document. The paper presents also results of experiments, which were carried out using the 20 News Groups and Reuters-21578 collections of documents and a collection of documents from an Internet portal of the Markiza broadcasting company.

It becomes a good habit to organize a data mining cup, a competition or a challenge at machine learning or data mining conferences. The main idea of the Discovery Challenge organized at the European Conferences on Principles and Practice of Knowledge Discovery in Databases since 1999 was to encourage a collaborative research effort rather than a competition between data miners. Different data sets have been used for the Discovery Challenge workshops during the seven years. The paper summarizes our experience gained when organizing and evaluating the Discovery Challenge on the atherosclerosis risk factor data.

In this paper we present dRAP-Independent, an algorithm for independent distributed mining of first-order frequent patterns. This system is based on RAP, an algorithm for finding maximal frequent patterns in first-order logic. dRAP-Independent utilizes a modified data partitioning schema introduced by Savasere et al. and offers good performance and low communication overhead. We analyze the performance of the algorithm on four different tasks: Mutagenicity prediction -- a standard ILP benchmark, information extraction from biological texts, context-sensitive spelling correction, and morphological disambiguation of Czech. The results of the analysis show that the algorithm can generate more patterns than the serial algorithm RAP in the same overall time.