|
|
Automated methods of textual content analysis and description of text structures
Chýla, Roman ; Smetáček, Vladimír (advisor) ; Rauch, Jan (referee) ; Uličný, Oldřich (referee)
Universal Semantic Language (USL) is a semi-formalized approach for the description of knowledge (a knowledge representation tool). The idea of USL was introduced by Vladimir Smetacek in the system called SEMAN which was used for keyword extraction tasks in the former Information centre of the Czechoslovak Republic. However due to the dissolution of the centre in early 90's, the system has been lost. This thesis reintroduces the idea of USL in a new context of quantitative content analysis. First we introduce the historical background and the problems of semantics and knowledge representation, semes, semantic fields, semantic primes and universals. The basic methodology of content analysis studies is illustrated on the example of three content analysis tools and we describe the architecture of a new system. The application was built specifically for USL discovery but it can work also in the context of classical content analysis. It contains Natural Language Processing (NLP) components and employs the algorithm for collocation discovery adapted for the case of cooccurences search between semantic annotations. The software is evaluated by comparing its pattern matching mechanism against another existing and established extractor. The semantic translation mechanism is evaluated in the task of...
|
|
|
Local and global analytical reports on results of data mining
Reischig, Zdeněk ; Rauch, Jan (advisor) ; Kliegr, Tomáš (referee)
Title: Local and global analytical reports on results of data mining Author: Zdeněk Reischig Department: Department of Software Engineering Supervisor: doc. RNDr. Jan Rauch, CSc. Supervisor's e-mail address: rauch@vse.cz Abstract: Thesis focuses on automatized support of local analytical reports creation, utilization of these reports as data sources for global analytical questions and creation of global analytical reports. In the thesis are suggested methods for comparison of rules. These methods are suitable for solving global analytical questions and can also help with composing of local analytical reports. Thesis also describes different kinds of background knowledge. One of them can be used for elimination of uninteresting rules or for finding data matrices with unusual relations. Other is necessary for solving global analytical questions, when rules are created over data matrices describing the same properties with different measures etc. Another important part of the thesis lies on providing a XML structure template for saving outputs of global analysis. It is also possible to use this XML structure for automatized generation of global analytical reports. Last part of the thesis is based on case study which shows how to use the guidelines and methods suggested in previous chapters. Case study...
|
|
|
Uživatelsky orientovaný jazyk pro řešení úloh DZD
Kováč, Michal ; Mrázová, Iveta (referee) ; Rauch, Jan (advisor)
The thesis discusses a new visual functional programming language and its use for data mining. The language is called Ferda and forms an integral part of the Ferda system, which is an application that has been created for data mining with the GUHA method. Functions of the language are represented by boxes. The source code is written as a connection of boxes; source files are project files of the Ferda system. The thesis describes the status of the Ferda system before this work from the point of view of the visual programming and describes possible enhancements to source files, then it presents a new basic set of boxes for the Ferda language and proposes other possible extensions of the language. Some of these proposals have been implemented as a part of this thesis. The last part includes examples of the use of the new language for data mining. One of these has also been implemented as a part of this thesis.
|
|
| |
|
|
Procedure SDKL-Miner for data mining
Tegze, Miron ; Lín, Václav (referee) ; Rauch, Jan (advisor)
A new data mining GUHA procedure SDKL-Miner is presented which extends current data mining system LISp-Miner developed on the VŠE Praha. The procedure mines for various patterns based on evaluation of two two-dimensional contingency tables. A simple comparing quantifier is implemented. Tests and new results for this quantifier are presented together with set of examples of it's use. New directions of future development are proposed. The use of statistical dependency tests and the use of both vector and matrix majorization are analyzed in more details. The procedure is a result of continued development of the academic LISp-Miner system for knowledge discovery and data mining.
|
|
|
User support 4ft-Miner procedure for data mining
Kupka, Daniel ; Ralbovský, Martin (referee) ; Rauch, Jan (advisor)
Up to now the file of knowledges with 4ft-Miner procedure was concerned only in the heads of data mining analysts. So, it was usefull, pursuant many analyses, to collect advices and experiences for data mining. This file of advices is also a part of this work. Furthermore, theoretical findings about using applicable 4ft-quantifiers according to practical situation are presented. This work also contains proposal of two new 4ft-quantifiers - above negation quantifier and pairing quantifier. For ASSOC-procedures users web application for collecting, editing advices, experiences and background knowledges is discribed and implemented. Benefit of this application is also in structuralization of inserted informations. The work also contains common design and the description of implementation of the tool for advices creation in Ferda DataMiner environement. Here, proceeding from possibilities ordered by their priority can be choosen. Informations provided by this tool are not only static, but they react on results of present and last analyses.
|
|
|
Classes of SD4ft-rules
Kodym, Jan ; Rauch, Jan (advisor) ; Holeňa, Martin (referee)
There are some classes of SD4ft-patterns defined in this work. SD4ft-patterns are the patterns which the GUHA procedure SD4ft-Miner deals with. The aim of this work was to define the classes of SD4ft-patterns, which have similar properties like the classes of association rules; in particular to find the criteria of correctness of deduction rules and data mining in databases, which contain an uncomplete information (NULL values). The logical calculi such that its formulae correspond to SD4ft-patterns and principles of work with uncomplete information are defined in this paper. Our effort was to utilize the existing knowledge of association rules (also called 4ft-patterns) to study SD4ft-patterns. Some potentially useful SD4ft-quantifiers based on useful 4ft-quantifiers are defined. The investigated classes of SD4ft-patterns were designed, so that they contain SD4ft-patterns based on these useful SD4ft-quantifiers.
|
|
|
Relational GUHA procedures
Kuzmin, Alexander ; Ralbovský, Martin (referee) ; Rauch, Jan (advisor)
The thesis belongs to "implementation category". The goal of this thesis is to design and implement relational extensions of the selected GUHA procedures that are implemented in LISp-Miner system. 4ft-Miner and SD4ft-Miner have been selected for the implementation of relational extensions. There have also been impemented several methods for attribute transformation. Procedures have been implemented in the Ferda Dataminer environment.
|
|
|
Experimental GUHA procedures
Kuchař, Tomáš ; Mrázová, Iveta (referee) ; Rauch, Jan (advisor)
The goal of this work is a new implementation of six GUHA procedures known from LISpMiner system (4ftMiner, SD4ft-Miner, CFMiner, SDCFMiner, KLMiner, SDKLMiner) into Ferda Data Miner system environment with respect to their futher research and development. GUHA procedure automatically generates patterns from user defined set of relevant patterns and tests if it is true in the analysed data. The output of the procedure consists of all prime patterns. The pattern is prime if it is true in the analysed data and if it does not immediately follow from the other more simple output patterns. Typical effective implementation of a GUHA procedure uses suitable database representation by bit strings. Tools were created for solving above-mentioned GUHA procedures. During works were extended options of entering the relevant questions set.
|
|
|
Using domain knowledge in applications of GUHA procedures
Ralbovský, Martin ; Mráz, František (referee) ; Rauch, Jan (advisor)
We presume for this work, that the GUHA procedures' abilities can be extended by using domain knowledge in a suitable form. There are two types of domain knowledge used in this work: background knowledge and ontologies. Background knowledge is a relatively unknown, yet perspective type of domain knowledge containing rules from domain experts in a written form. The work describes this knowledge and introduces suitable rule formalization. Furthermore, a tool for rule validation is described and implemented. The implementation is done in the Ferda environment, a modular environment for visual GUHA data mining. The work also describes first practical experience with this tool. Concerning domain ontologies, the work greatly extends up-to-date knowledge of usage of ontologies in the GUHA data mining. There are all the former techniques for ontology usage mentioned and extended; in suitable cases the implementation design in Ferda environment is presented. The results of the work are discussed with respect to further usage in the Ever-Miner system, a next generation knowledge discovery tool
|
guest ::
