translated by Google

Machine-translated page for increased accessibility for English questioners.

Program of seminars for 2003/2004

St. 12.00 - 13.00, A107
Seminar leader: doc. Prof. RNDr. Ivan Kopecek, CSc. ( )

Autumn 2003

September 24, 2003
Introductory seminar of the autumn semester
Information about seminar concept in the autumn semester.
Seminar program agreement.
1. 10. 2003
Jan Flasar
Local area interaction
In this presentation we will focus on interaction in virtual space - we will only talk about methods for selecting and manipulating objects. We will show you the problems we can encounter here and how to solve them. Then we look at the design methods that should accelerate user interaction with virtual objects. And we also wonder where these techniques can be used.
8. 10. 2003
Michal Batko
Scalable and Distributed Data Structures
Since the emergence of high-speed networks, interest in distributed systems increased considerably. Modern computer networks have large enough bandwidth, so that it is more expensive for the application to access local disk, then it is accessed by the RAM of another computer on the network. We can therefore distribute data among network nodes using a distributed system. For efficient use of such systems, new techniques for storage, access and processing of distributed data. Scalable Distributed Data Structures (SDDS) represent a step towards efficient storage and access to distributed data, since they are well suited for large scale distributed systems. Distributed techniques can also be used to enhance the processing of huge datasets of objects, where we can compute similarities between those objects (examples can be image recognition, audio samples matching, molecules searching, ...). One method to accomplish this task is to use "metric space" abstraction for classes of objects that allow defining and "distance" function. This method is used in scalable distributed data structure GHT *, which allows to perform similarity queries on generic metric spaces in distributed environment.
15. 10. 2003
Prof. Manfred Kudlek
University of Hamburg
Cyclic Words
Cyclic words play and certain role in DNA computing. Several methods to generate languages of cyclic words are presented. One is taking cyclic permutations of classical languages, another one to solve algebraic systems of equations on cyclic words. The relations between the classes are defined in this way.
October 22, 2003
D. Šafránek
Visual Specification of Concurrent Systems
October 29, 2003
P. Motlíček
Modeling Spectra and Temporal Trajectories in Speech Processing
This work investigates the application of spectral and temporal processing algorithms developed for feature extraction in Automatic Speech Recognition (ASR) and for very low bit-rate speech coding. In the first part, various spectral processing techniques are investigated for robust parameterization of speech. We are especially focused on all-field modeling based techniques that use, as the major processing block, the suppressive speaker-dependent model in the auditory spectrum. Such techniques that use the model spectrum are advantageous as opposed to directly using signal audit spectrum. Further, the work investigates the inclusion of temporal information as an additional speech parameter in ASR. We primarily concentrate on the use of frequency-localized temporal trajectories (patterns). In the last part, the use of spectral and temporal processing techniques for Very Low Bit Rate (VLBR) speech coding is investigated. We study the properties of authoritative model-based features employed in speech coder. Our coding approach is based on a proper selection of speech units automatically derived from the Automatic Language Independent Speech Processing (ALISP) tool.
5. 11. 2003
P. Cenek
Dialogue System Design Abstract:
Design of dialogue systems, ie systems that use natural language for communication with users, is specific in many ways. The structure of such systems will be introduced and particularities of their design will be discussed. Examples of applications will be presented.
F. Procházka
Knowledge representation and acquisition via UIR system and identification expressions
The presentation will introduce the principle of knowledge representation and acquisition via the UIR (Universal Information Recorder) system, UIR ontology and called identification expressions. Using this method enables us to specify knowledge base as well as the inference machine of information / knowledge-based system on a high conceptual level of abstraction. This method will be compared with the method based on Transparent Intensional Logic (TIL). The UIR system is used in real commercial applications.
12.11. 2003
M. Hrad
Time Stamping
In order to use digital documents as a medium for keeping information it is necessary to apply a technique which assures the integrity of digital documents over a long time period. What is needed is a method of time-stamping digital documents that can time-stamp the date itself and make a stamp and document with a time that is different from the actual one. Time Stamping is a technique that proves the existence of certain digital data to a specific point in time. The presentation will focus on present situation, simple and linking schemes, requirements of TSA and practical realization.
19.11. 2003
V. Kovalčík
Occlusion Culling
Despite the fact that the processors are faster and more complex features are added to the graphic cards every year, complicated visual scenes can not be displayed fast enough yet. Thus some new searching methods for visible objects in huge scenes are needed. In this talk some of them will be discussed and the directions of the research in this area will be presented.
26.11. 2003
P. Holub
TCP and beyond: High-bandwidth long-distance networks
Bandwith provided by current high-speed networks grow even faster than computing and storage capacity. Some other properties, however, can not be improved at such times as due to limitations imposed by the laws of physics. Most important is latency, influencing dramatical behavior of TCP - the most common Internet protocol for reliable transfers. Building on basic ideas described by Sven Ubik during his talk two weeks ago I will introduce improvements to TCP protocol designed to cope with networks described above. We will examine approaches different from one used by TCP as well.
3.12. 2003
J. Thomas
English: couldn't give a damn
Together we will assemble a picture of the best speakers of English, how they came to be, and consider who can join their ranks. The question is: can anyone be a "best speaker of English", and if so, what does it take? Conversely, we will consider what it takes to ensure eternal exclusion from the club.
A website for those who are interested can be found at Note: comments on pronunciation problems can be found (underscore before 'pron').
10.12. 2003
P. Kolčárek
Level of detail: Haptics
We will present a possible way of using LOD in a haptic environment. Level of detail is a technique commonly used in computer graphics for displaying complex 3D models. However, it can be used in other fields as well. We will discuss some advantages and drawbacks of using LOD in graphics and haptics. A brief outline of the LOD algorithm in haptics will be presented at the end.
17.12. 2003
Poster Session
The posters created during the autumn semester will be presented.

Spring 2004

25. 2. 2004
Spring semester introductory seminar
Information about seminar concept in the autumn semester.
Seminar program agreement.
3. 3. 2004
Luděk Bártek
Generating dialog interfaces
Nowadays, there is a boom in Internet applications based on W3 consortium standards. Unfortunately, most of these applications have only a graphical output, which is not suitable for people with different needs, for access by various thin clients based on the voice. This lecture can describe the issue.
March 10, 2004
Jan Kohout
Precitipation data processing
March 17, 2004
James Thomas
Help yourself to help yourself
This talk will follow on from my December one in which I attempted to make much of the conscious and subconscious choices that are involved in producing language and in particular foreign language, whether spoken or written. The parallel thrust in that talk of our need to equip ourselves with as many linguistic resources as possible to make the most reasoned choices. This was a call to learn more, pursue studies. A brief illustration was also given by how corpora can assist in our decision making. In this talk, I would like to develop this theme. General corpora can answer general questions, which is important since some general aspects of language. But focusing on the characteristics of specific language tasks, writing academic papers for example, it is advantageous to consult and corpus of the specific genre you are working in. And if such a corpus does not exist, we must create one: the Brno Informatics Corpus, perhaps!
March 24, 2004
Miroslav Křipač
Data Sharing Within Database Clusters
Database clusters meet current demand for scalability and high availability of mission critical systems. However, conventional implementations of single-purpose shared disk subsystems limit their wide-spread adoption primarily for cost reasons. This presentation will focus on an alternative software approach to data sharing methods and its impact on system performance.
March 31, 2004
L. Svoboda
Question Answering Systems
At present, the Internet makes available many databases, each with its own interface (weather, traffic links, currency rates, ...). The natural endeavor is to create one universal interface that allows for natural language communication. Basic principles and current state of development in the field of answering questions are presented.
April 7, 2004
V. Kadlec
Syntactic Analysis of Natural Languages
The libkp} comparison and evaluation system for the syntactic processing of natural languages. The analysis of natural language sentences is based on context-free grammar for given language. The internal representation of the derivative trees should allow for the application of contextual constraints, eg the fulfillment agreement. In general, this is a NP-complete problem. That is why it is required to develop new evaluation algorithms. Then we will be able to work effectively with the trees, where these constraints were applied. We can also select the most probable trees based on the frequency characteristics obtained from tree-banks.
The system is language-independent, however, it is optimized for large and highly ambiguous grammars for Czech (thousands of rules). We also expect this tool to be used for comparing and evaluating different CFG-parsing algorithms.
14. 4. 2004
M. Marciniszyn
Bipartite Entanglement in Mixed States
A new approach, which considers quantum entanglement as a source for the transmission of information, is presented in the talk. The algorithm of computation of our measure is polynomially slower in comparison to those by Bennet et. al., however, this drawback is fully compensated by the possibility of measuring mixed states entanglement. The approach is presented from the point of view of computer scientists, no knowledge of quantum mechanics is supposed.
April 21, 2004
J. Stibor
Interphase Chromosomes Segmentation
The process of chromosomes segmentation and their localization is essential for genetic diseases diagnostic (Down Syndrome, leukemia, ...). This talk will explain the method of interphase chromosomes segmentation. The method is based on Voronoi Tesselation of 2D and 3D images. The duality with Delaunay Graph is used in the implementation of this method.
April 28, 2004
J. Hejtmanek
Distributed Data Storage
5. 5. 2004
R. Pelanek
Checking Model - An Overview
Model checking is a formal method aimed at design and verification of safety-critical systems. The talk will provide an overview of this technique. We will discuss the idea of the approach; main problems and remedies; History, state of the art and future of the field. Finally, we will mention the relevant research in the ParaDiSe laboratory.
May 12, 2004
P. Adamek
Effective Evaluation of XPath Expressions
XPath is the most common language for searching XML documents. This paper describes the issue of XPath expression evaluation and outlines some methods that use indices for speeding up this process. In addition, the requirements and benefits of each method are discussed; great emphasis is put on the desired abilities of indices.
S. Barton
Designing Indexing Structure for Discovering Relationships in RDF Graphs
Discovering the complex relationships between entities is one way of benefiting from the Semantic Web. This paper discusses new approaches to incorporating rho-operators into RDF querying engines, which will enable viable viability. Such incorporation is based on creating an index that describes the original RDF graph. The index is created in two steps. Firstly, it transforms the RDF graph into a forest of trees and then creates an extended signature for each of these trees. The signatures are accompanied by additional information related to the transformed problematic nodes breaking the tree structure.

Jan Flasar
Pavel Kolcarek
Radka Pospisilova
Lukas Kokrment
Vaclav Kadlec
Vit Kovalcik
Pavel Cenek
Michal Batko
Tomas Hnilica