translated by Google

Machine-translated page for increased accessibility for English questioners.

Seminar program for 2001/2002

St. 12.00 - 13.00, A107
Lecturer: doc. RNDr. Ivan Kopeček, CSc. ( )
4. 10. 2001
Introductory Seminar Seminar
Information on Seminar Concept in the Autumn Semester.
Agenda of the seminar.
11. 10. 2001
James Edward Thomas, BA
Natural Language Processing for learners of foreign languages: some problems and suggested solutions.
When someone is learning a foreign language, a great amount of high quality language input is needed. This is especially true when discovery learning is encouraged. When students can discover for themselves some elements of how language works, much deeper processing is taking place. This in turn has long-term benefits due to a more conceptual grasp of the language that has been derived from the language being studied rather than through mere surface translations. Non-native speakers do not have the luxury of time to be exposed to a great amount of high quality input, on which they might otherwise test language hypotheses. When students have a large amount of high quality language in a database through which they can isolate specific language items, they can observe, hypothesize and test their findings. And this is what Data Driven Learning is - a pedagogical application of NLP. I am currently working on a small program that can present lines of text in order of the sum of the items in each example. The result should be a screen of text, with the target item centered and its context presented in order of familiarity of vocabulary items. For research purposes, this concordancer is all but ready. It remains to find out if the hypothesis that providing students with contexts is likely to understand contributes to the intended pedagogical gain.
RNDr. Milan Drasil, CSc.
Introduction to Geographic Information Systems
-The basic orientation of the lecture.
- What is a geographic information system, the subject part of a geographic information system.
-GIS concepts, maps of the Czech Republic.
-Space data structures, vector space representation, raster space representation, topological relationships in spatial data.
-Efficient methods of access to spatial data, definition of general spatial query, primary and secondary spatial query filter, spatial indexing methods, database context.
-Operation over vector spatial data.
-Graphic image processing operations usable in GIS.
-Topological Tasks in GIS.
18 October 2001
J. Pokluda
IA-64 (Intel Architecture 64 bit)
Intel and Hewlitt-Packard teamed up to create a new 64-bit architecture that is the successor of the RISC and CISC architectures. They called it EPIC (Explicitly Parallel Instruction Computing). This architecture could serve as a platform for creating a processor that can run in parallel, with more simplified hardware requirements (more than CISC) to include more advanced drives to achieve better performance. Compared to the previous Intel processor, they will contain a large amount of registry. Zaroven should maintain consistency with IA-32, both hardware and PA-RISC software.
The lecturer discusses the main ideas behind this architecture.
25. 10. 2001
Robert Batůšek
Corpus-oriented speech synthesis
Traditional speech synthesis in the time domain uses (relatively) a small number of usually phonetically motivated units (allophytes, dipony, syllables) from which the collection is traced back. In the synthesis, however, the segments are deformed as a result of prosodic modifications, which degrades the quality of the resulting speech. Therefore, more modern approaches make use of a greater number of heterogeneous segments (ranging from 10 ^ to 10 ^ 6). With such a record database, the need for prosodic modifications to the background subsides, but it is an interesting problem to select units that are in some sense optimal.
1 November 2001
T. Capek
Enhancing Czech WordNet
After explaining how WordNet and its follower EuroWordNet are and how they came about, I would like to focus on the current state of the EuroWord Net at our faculty and look at its expansion. To do this, I will use the experience I have gained in a specific work on spreading a cluster of lexical data and setting up the problems associated with this work.
8.11. 2001
Miloslav Nepil
Theory Refinement in Active Learning: An Application in Morphological Disambiguation of Czech "
Our work stands in the border between machine learning and natural language processing. We employ inductive logic programming to induce a set of rules for a morphological disambiguation of Czech. In this task, it is expensive to collect a sufficient amount of labelled examples needed for training. On the other hand, unlabelled texts are freely available. The objective of active learning is to select those examples that will be most informative for the learner. Since this example selection proceeds step-by-step, it is possible to refine the actual theory in order to disambiguate the new selected examples correctly. Thus, we aim at a combination of techniques: theory refinement within active learning. The outline of our approach and some experimental results will be presented.
15. 11. 2001
Eva Hladká
Videoconference in the research project High-speed National Research Network and its new applications
Research Plan The high-speed national research network and its new applications is the most important project of the CESNET association and it operates and develops the CESNET2 network. The aim of the presentation will be to familiarize students with the group of application projects with videoconferencing themes. The first part will briefly describe all videoconferencing projects, issues that are solved and the results achieved. The second part will be devoted to WAN videoconferencing project, which is solved at FI and ÚVT MU, especially the academic video conferencing server and the systems available on the server. Experience with H.323 technology will be shown at Megaconference III (Internet2 meeting), which took place in October 2001. Finally, the use of streaming servers for multimedia transmissions will be mentioned.
22. 11. 2001
F. Procházka
Design of tools for universal cognitive modeling
The concept of the Tonal tool for universal cognitive modeling, on whose design and implementation I work, will be presented. The introduction briefly mentions the issue of finding a suitable conceptual device for cognitive modeling.
Next, the theoretical background of the Tonal instrument concept. I will divorce the context with the natural language semantics and Transparent Intentional Logic (TIL). Describe the modeling entity of the Tonal tool and on the examples I will illustrate their use. We mention the possibility of using the model of access to the field of artificial intelligence (machine learning, data mining, neuron site, genetic algorithms).
The general description of the dynamics of cognitive models is based on the model of the transfer of attention in the model.
On top of that, I will introduce the role of Tonal as a workbench for research in the field of cognitive science and its use in computer science, which is forced by a brother to research the "cognitive element" in thought.
29. 11. 2001
P. Sojka
Contextual text processing
6.12. 2001
Petr Matula
Image registration in fluorescence microscopy
13 December 2001
R. King
Semantic desambiguation of the pathins
WSD or Jaky Mean Me? * The goal of the WSD system is to determine the appropriate meaning of the words in the text. We would like to introduce the best known techniques and present experiments. * Basic word-type algorithms. Slovenius pristupy. Metrics. Evaluation. * Use cluster analysis. Comparison of SAs on verbal forms and SAs on contexts represented in vector space.
R. Sedláček
Morfematically analyze the casein
The aim of the lecture is to present ideas for the design and implementation of a morphematic analyzer based on a combination of form-forming and word-formation analysis. As the shape-forming analysis has already been algorithmically described, one step is to smooth the morphematic analysis, namely algorithmization of the word-word analysis. The result of the morphematic analysis is to determine the structure of the analyzed word and the identification of the individual morphs. The aim is to find a suitable representation of the morphus and to evaluate the possibilities of automatic processing of selected derivation processes. The application of the morphemic analyzer on the corpus of Czech texts will then yield usable statistical data reflecting the current state of the cousin.
20. 12. 2001
Presentations will be created in the autumn semester.

Spring 2002

27. 2. 2002
Introductory Seminar of the Spring Semester
Information about the seminar in the spring semester.
Agenda of the seminar.
6.3. 2002
Jiri Barnat, Jan Strejcek, David Safranek
Introduction to system verification (model checking)
In the informative age, our lives are sometimes bound to the flawlessness and reliability of computer systems. These systems are becoming increasingly complex and complex. This implies the need for automated methods of verifying the authenticity of the mentioned systems. The lecture will focus primarily on one of these methods, the so-called model checking. The method will be adequately explained and demonstrated on a pseudo-real system.
David Safranek: Formal Verification Methods. Modeling and system specifications. Definition of a modeling check.
Jan Strejcek: Modeling assistant for automation theory.
Jiri Barnat: Presentation of the model of check on the pseudo-real system of drawing, realized by RCX 2.0 micropocity. SPIN will be used for verification.
13.3. 2002
Jan Pazdziora, Michal Brandejs, Iva Hollanova, Mirka Misakova
Introduction to system verification (model checking)
The lecture deals with the support of decision-making and management processes and information flows using university information systems (UIS - University Information System) in connection with the development of computer networks and possibilities of remote work with data. This in turn leads to changes in the sets of typical users of each application, and to a significant increase in the number of new users, in the case of the University of Teachers and students. When deploying the information system, one of the important input decisions is the degree of involvement of affected persons in direct work with the system.
20. 3. 2002
V. Dohnal
Multimedia Data Indexing
Recently, the need to manage various types of data stored in large computer repositories has drastically increased and resulted in the development of multimedia database systems with the goal of uniformly managing voice, video, image, text and numerical data. Among the many research challenges that multimedia technology involves - including content placement, presentation, synchronization, ... - content-based retrieval plays a dominant role. In particular, there is an urgent need for indexing techniques that can support the execution of similarity queries. This paper presents some basic index structures.
P. Cuba
Support Vector Machines
Support Vector Machines (SVM) is a fairly new method that has recently gained increasing popularity in addressing database search tasks such as classification, regression, and more. During this lecture, the basic principles of SVM will be explained from the geometric view. It will also explain the possibility of applying the method to different types of data (eg structured data) and mentioning applications to real data.
27.3. 2002
L. Bartek
Dialogue Interface Generator
The dialog interface seems to be an appropriate solution for communicating the user with the computer. This solution will allow the number of potential users of computer systems to be captured by people who do not know what to do with these systems. One of the dialog entry standard used in the dialog interfaces is VoiceXML. Due to the large number of existing applications whose user interface is based on HTML, the idea was to automatically generate dialogs for the above mentioned applications. In the lecture, lectures will be introduced to the basic principles of dialog generation. Further, the problematic points of this approach will be discussed, along with the suggestion of possible solutions.
3.4. 2002
Petr Matula
Registration of images taken with an axial tomograph
The Optical Microscopy Laboratory (LOM) has an interesting device available at the patented University of Heidelberg, which allows observing objects under a microscope (a cell) from a multi angle - axial tomograph. The cells are attached to the fibers that can be rotated. The axial resolution of the microscope is always superior to lateral resolution. Compositions of several images can be improved. A major step towards improving resolution is the good recording of the image (ie, the relative picture of the image so that the observed objects are as close as possible). During the lecture will be introduced an effective algorithm for the registration of the point set points developed in LOM, which is suitable for this purpose.
10.4. 2002
Pavel Gaura
Reading mathematical expressions to the blind
Visually impaired students currently have the opportunity to acquire knowledge faster and to a greater extent thanks to the development of computer technology and information technology. However, for the study of graphic objects, they still need the assistance of a visible worker who adjusts materials to the appropriate form. This category also includes the study of mathematical texts. The lecture will provide a brief overview of how math texts can be passed on to blind people. We will focus on different ways of reading or editing mathematical expressions. Furthermore, students will be introduced to the design of a multimodal way of reading mathematical expressions and using a dialog for their editing. In this context, the MathML markup language will be mentioned. At the end of the lecture will be demonstrated the prototype system for the study of mathematical expressions by the blind.
17.4. 2002
David Antoš
Pattern recognition in computer typography
We describe the pattern technique as a means of obtaining and recognizing information in the data, and using the word-cutting patterns in the slot machine at the same time. We have developed the PatLib library for general pattern manipulation, and we intend to use it more widely for recognizing various phenomena in quote material, such as sentence boundaries, segmentation, compound word seams. We will describe the architecture of the library. Next, we formulate the idea of ​​the Pattern Translation Process, a pattern-recognition machine, on a stream of expanded rate material. We discuss the benefits that PTP would bring over "traditional" techniques (such as Omega Translation Processes).
24.4. 2002
Pavel Matula
Reconstruction of round objects using simplex networks
The lecture will address the problem of how to determine and represent the surface of an unknown real object based on the set of input points measured on its surface. An overview of the methods that are used to resolve this issue will be provided. Since we observe mainly round objects (cell nuclei) in the optical microscopy laboratory, we will focus on the discussion of individual methods towards usability in our application. Deeper development of the method based on deformation of the simplex network, which is suitable for the reconstruction of round objects, will be introduced.
15.5. 2002
Aleš Borek
22 May 2002
The posters created in the spring semester will be presented.