February 2020 - informatik-vortraege

Einladung Informatik-Oberseminar Malte Nuhn
by Sekretariat I6 30 May '22

30 May '22

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 12. Juli 2019, 10.00 Uhr Ort: Informatikzentrum, E3, Raum 9222 Referent: Dipl.-Inform. Malte Nuhn Thema: Unsupervised Training with Applications in Natural Language Processing// Abstract: The state-of-the-art algorithms for various natural language processing tasks require large amounts of labeled training data. At the same time, obtaining labeled data of high quality is often the most costly step in setting up natural language processing systems.Opposed to this, unlabeled data is much cheaper to obtain and available in larger amounts.Currently, only few training algorithms make use of unlabeled data. In practice, training with only unlabeled data is not performed at all. In this thesis, we study how unlabeled data can be used to train a variety of models used in natural language processing. In particular, we study models applicable to solving substitution ciphers, spelling correction, and machine translation. This thesis lays the groundwork for unsupervised training by presenting and analyzing the corresponding models and unsupervised training problems in a consistent manner.We show that the unsupervised training problem that occurs when breaking one-to-one substitution ciphers is equivalent to the quadratic assignment problem (QAP) if a bigram language model is incorporated and therefore NP-hard. Based on this analysis, we present an effective algorithm for unsupervised training for deterministic substitutions. In the case of English one-to-one substitution ciphers, we show that our novel algorithm achieves results close to human performance, as presented in [Shannon 49]. Also, with this algorithm, we present, to the best of our knowledge, the first automatic decipherment of the second part of the Beale ciphers.Further, for the task of spelling correction, we work out the details of the EM algorithm [Dempster & Laird + 77] and experimentally show that the error rates achieved using purely unsupervised training reach those of supervised training.For handling large vocabularies, we introduce a novel model initialization as well as multiple training procedures that significantly speed up training without hurting the performance of the resulting models significantly.By incorporating an alignment model, we further extend this model such that it can be applied to the task of machine translation. We show that the true lexical and alignment model parameters can be learned without any labeled data: We experimentally show that the corresponding likelihood function attains its maximum for the true model parameters if a sufficient amount of unlabeled data is available. Further, for the problem of spelling correction with symbol substitutions and local swaps, we also show experimentally that the performance achieved with purely unsupervised EM training reaches that of supervised training. Finally, using the methods developed in this thesis, we present results on an unsupervised training task for machine translation with a ten times larger vocabulary than that of tasks investigated in previous work. Es laden ein: die Dozentinnen und Dozenten der Informatik _______________________________________________ -- -- Stephanie Jansen Faculty of Mathematics, Computer Science and Natural Sciences HLTPR - Human Language Technology and Pattern Recognition RWTH Aachen University Ahornstraße 55 D-52074 Aachen Tel. Frau Jansen: +49 241 80-216 06 Tel. Frau Andersen: +49 241 80-216 01 Fax: +49 241 80-22219 sek(a)i6.informatik.rwth-aachen.de www.hltpr.rwth-aachen.de Tel: +49 241 80-216 01/06 Fax: +49 241 80-22219 sek(a)i6.informatik.rwth-aachen.de www.hltpr.rwth-aachen.de

3 15

Einladung: Informatik-Oberseminar Gereon Kremer
by Gereon Kremer 28 Feb '20

28 Feb '20

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Donnerstag, 12. März 2020, 10:45 Uhr Ort: Raum 9222, Gebäude E3, Ahornstr. 55 Referent: Gereon Kremer, M.Sc. Theory of Hybrid Systems Thema: Cylindrical Algebraic Decomposition for Nonlinear Arithmetic Problems Abstract: We explore the usage of the cylindrical algebraic decomposition method for satisfiability modulo theories solving, both theoretically and experimentally. This method is commonly understood as an almost atomic procedure that gathers information about an algebraic problem and then allows to answer all kinds of questions about this algebraic problem afterwards. We essentially break up this method into smaller components that we can then process in varying order to derive the particular piece of information – whether the problem is satisfiable or unsatisfiable – allowing to avoid some amount of computations. As this method often exhibits doubly exponential running time, these savings can be very significant in practice. We then turn to an alternative approach to satisfiability modulo theories solving commonly called model-constructing satisfiability calculus. The core idea of this framework is to integrate the theory reasoning, in particular the construction of a theory model, very tightly with the Boolean reasoning. The main theory reasoning engine is again based on the cylindrical algebraic decomposition method, though we focus more on the overall framework here. Es laden ein: die Dozentinnen und Dozenten der Informatik -- Gereon Kremer Lehr- und Forschungsgebiet Theorie Hybrider Systeme RWTH Aachen Tel: +49 241 80 21243

1 0

Einladung Informatik-Oberseminar Christian Corsten
by Corsten, Christian 25 Feb '20

25 Feb '20

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Dienstag, 10. März 2020, 9.30 Uhr Ort: Raum 2222, Ahornstr. 55 (Informatikgebäude, Hauptbau; Seminarraum i10) Referent: Christian Corsten, M.Sc. Lehrstuhl Informatik 10 Thema: Use the Force: How Force Touch Improves Input on Handheld Touchscreens Abstract: Handheld devices, such as smartphones, have become essential tools in our everyday life. We use them, e.g., to contact people, browse the web, or take pictures. For whatever use, to interact with the handheld device, we hold it with one or two hands and touch with our fingers on the built-in touchscreen. However, this interaction is often constrained to simple contact between the finger and the flat display glass, although touch offers further, richer properties. One such rich property is the intensity of a touch, i.e., its force, that the user applies with every tap to the touchscreen. Incorporating this property into the user’s interaction with the handheld device enables her to become more expressive with every single touch. In this thesis, we present a series of interaction techniques that take advantage of force touch input to make handheld interaction more efficient: When holding the device with two hands in landscape orientation, most of the fingers are unavailable for interaction, since they rest at the back of the device (BoD), holding it in place. Using BoD force input, we can make efficient use of these fingers without sacrificing stability of the device grip. Our technique, BackXPress, enables quick access to shortcuts and menus to augment users’ touch interaction with the frontal screen. For single-handed device use, users can only use their thumb to interact with the frontal touchscreen but cannot reach everywhere without re-grasping the device. Our virtual thumb extension, ForceRay, lets the user cast a ray at unreachable targets and control a cursor on that ray that moves closer to these targets the more force is applied. The technique is ergonomic for the thumb and enables users to maintain a steady device grip. Targets located at the screen edges, like menus and navigation buttons, are acquired quickly. Selection of values from long ordered lists, such as picking a date or time, can also be sped up when using force input. With our Force Picker, users scroll through the value range at various speeds, with the speed being coupled to the force exerted by the thumb. Prior rolling of the thumb to the left or right sets the scrolling direction. Compared to touch-based pickers, our Force Picker not only makes selection faster, but also only consumes little screen space since the gesture footprint for force input is much smaller. While controlling force via fingers requires practice, we show that with training and algorithmic optimizations, users become quickly familiar with force input and gain the benefits of the added expressiveness for handheld interaction. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0

Einladung: Informatik Oberseminar Daniel Zielasko (Raumänderung!)
by Daniel Zielasko 17 Feb '20

17 Feb '20

+********************************************************************** * * * Einladung * * * * Informatik-Oberseminar * * * +********************************************************************** Zeit: Freitag, 21. Februar 2020, 9.00 Uhr Ort: IT Center Erweiterungsbau, Kopernikusstraße 6, S003 Referent: Daniel Zielasko, M.Sc. Lehrstuhl Informatik 12 Thema: DeskVR: Seamless Integration of Virtual Reality into Desk-based Data Analysis Workflows Abstract: In this work, we are looking into the possibilities and unique challenges virtual reality today offers for (office) desk-based scenarios, as they are ubiquitous in data analysis. We characterize the scenario, introduce the term deskVR, and name the technical challenges that come with it. Furthermore, we tackle specific demands in two pillars of interaction in virtual reality, selection & manipulation, and navigation. Then, we investigate passive and active methods to prevent and reduce cybersickness, as for us, tackling cybersickness is one if not the most critical tasks that have to be solved to integrate virtual reality into everyday life successfully. Finally, we apply the methods and findings made in this thesis to a prototypical application framework for immersive 3D graph exploration, serving as proof of concept for the integrability of virtual reality into desk-based working scenarios. In the graph visualization domain, we then also propose new vertex positioning and edge bundling methods that address challenges arising with the performed up-projection into 3D interactive space. Es laden ein: die Dozentinnen und Dozenten der Informatik

1 0