Henryk Rybinski

Intelligent Tools for Building a Scientific Information Platform

This book is a selection of results obtained within two years of research per- formed under SYNAT - a nation-wide scientific project aiming at creating an infrastructure for scientific content storage and sharing for academia, education and open knowledge society in Poland. The selection refers to the research in artificial intelligence, knowledge discovery and data mining, information retrieval and natural language processing, addressing the problems of implementing intelligent tools for building a scientific information platform.This book is a continuation and extension of the ideas presented in Intelligent Tools for Building a Scientific Information Platform published as volume 390 in the same series in 2012. It is based on the SYNAT 2012 Workshop held in Warsaw. The papers included in this volume present an overview and insight into information retrieval, repository systems, text processing, ontology-based systems, text mining, multimedia data processing and advanced software engineering.

Finding Reducts in Composed Information Systems

A set-theoretical approach to finding reducts of composed information systems is presented. It is shown how the search space can be represented in form of a pair of boundaries. It is also shown, how reducts of composing information systems can be used to reduce the search space of the composed system. Presented solutions are implied directly from the properties of composed monotonic Boolean functions.

Reducing Information Systems with Uncertain Attributes

We present Rough Set approach to reasoning in information systems with uncertain attributes. A similarity relation between objects is introduced. The relation is a tolerance relation. A reduction of knowledge we propose eliminates only information, which is not essential from the point of view of classification. Our approach is general in the sense it does not assume anything about the semantics of null values and uncertain multivalued attributes. We show how to find decision rules, which have minimal number of conditions and do not increase the degree of non-determinism of the original decision table.

Dataless Transitions Between Concise Representations of Frequent Patterns

For many data mining problems in order to solve them it is required to discover frequent patterns. Frequent itemsets are useful e.g. in the discovery of association and episode rules, sequential patterns and clusters. Nevertheless, the number of frequent itemsets is usually huge. Therefore, a number of lossless representations of frequent itemsets have recently been proposed. Two of such representations, namely the closed itemsets and the generators representation, are of particular interest as they can efficiently be applied for the discovery of most interesting non-redundant association and episode rules. On the other hand, it has been proved experimentally that other representations of frequent patterns happen to be more concise and more quickly extractable than these two representations even by several orders of magnitude. Hence, such concise representations seem to be an interesting alternative for materializing and reusing the knowledge of frequent patterns. The problem however arises, how to transform the intermediate representations into the desired ones efficiently and preferably without accessing the database. This article tackles this problem. As a result of investigating the properties of representations of frequent patterns, we offer a set of efficient algorithms for dataless transitioning between them.

Data mining in incomplete information systems from rough set perspective

Mining rules is of a particular interest in Rough Sets applications. Inconsistency and incompleteness issues in the information system are considered. Algorithms, which mine in very large incomplete information systems for certain, possible and generalized decision rules are presented. The algorithms are based on efficient data mining techniques devised for association rules generation from large data bases. The algorithms are capable to generate rules both supported by the system directly and hypothetical. The rules generated from incomplete system are not contradictory with any plausible extension of the system.

Discovering Compound and Proper Nouns

The identification of appropriate text tokens (words or sequences of words representing concepts) is one of the most important tasks of text preprocessing and may have great influence on the final results of text analysis. In our paper, we introduce a new approach to discovering compound nouns, including proper compound nouns. Our approach combines the data mining methods with shallow lexical analysis. We propose a simple pattern language for specifying grammatical patterns to be satisfied by extracted compound nouns. Our method requires annotating the words with part of speech tags, thus to this extent, it is language-dependent. Based on the data mining GSP algorithm, we propose T-GSP as its modification for extracting frequent text patterns, and in particular, frequent word sequences that satisfy given grammatical rules. The obtained sequences are regarded as candidates for compound nouns. The experiments have proven very high quality of the method.

Computation of reducts of composed information systems

A set-theoretical approach to finding reducts of composed information systems is presented. It is shown how the search space can be represented in form of a pair of boundaries. It is also shown, how reducts of composing information systems can be used to reduce the search space of the composed system. Presented solutions are implied directly from the properties of composed monotonic Boolean functions.

Text onto miner: a semi automated ontology building system

This paper presents an overview of the results of the project undertaken by the Warsaw University of Technology Institute of Computer Science as a part of research agreement with France Telecom. The project goal was to create a set of tools - both software and methods, that could be used to speed up and improve a process of creating ontologies. In the course of the project a new ontology building methodology has been devised, new text mining algorithms optimized for extracting information useful for building an ontology from text corpora have been proposed and an universal text mining toolkit - TOM Platform - have been implemented.

FARICS: a method of mining spatial association rules and collocations using clustering and Delaunay diagrams

The paper presents problems pertaining to spatial data mining. Based on the existing solutions a new method of knowledge extraction in the form of spatial association rules and collocations has been worked out and is proposed herein. Delaunay diagram is used for determining neighborhoods. Based on the neighborhood notion, spatial association rules and collocations are defined. A novel algorithm for finding spatial rules and collocations has been presented. The approach allows eliminating the parameters defining neighborhood of objects, thus avoiding multiple “test and trial” repetitions of the process of mining for various parameter values. The presented method has been implemented and tested. The results of the experiments have been discussed.

Discovering Synonyms Based on Frequent Termsets

Synonymy has been of high importance in information retrieval and automatic indexing. Recently, in the view of special needs for domain ontology building and maintenance, the problem returns with a higher demand. In the presented paper, we present a novel text mining approach to discovering synonyms or close meaning terms. The offered measures of closeness of terms (or their contexts) are expressed by means of data mining notions; namely, frequent termsets and association rules. The measures can be calculated by using data mining techniques, such as the well known Apriori algorithm. The approach is domain-independent and large-scale. It is, however, restricted to the recognition of parts of speech. In that sense the approach is language dependent, up to the language dependency of the parts of speech tagging process. The experimental results obtained with the approach are presented.