+**********************************************************************
*
*
* 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
+**********************************************************************
*
*
* Einladung
*
*
*
* Informatik-Oberseminar
*
*
*
+**********************************************************************
Zeit: Dienstag, 10. Dezember 2019, 12:00 Uhr
Ort: Raum 9222, Gebäude E3, Ahornstr. 55
Referent: Matthias Hoelzel, M.Sc.
LuF Math. Grundlagen der Informatik (Lehrgebiet Informatik 7)
Thema: Fragments of Existential Second-Order Logic
and Logics with Team Semantics
Abstract:
Team semantics is the modern basis for logics of dependence and
independence whose formulae are not evaluated with single assignments,
but with sets of such assignments. We study different fragments of
logics with team semantics and existential second-order logic (ESO) such
as the union-closed fragments of these logics, inclusion logic of
restricted arity and logics with team semantics using weaker dependency
concepts.
We will present syntactic characterisations for the union-closed
fragment of ESO and inclusion-exclusion logic. In order to obtain these
characterisation results, novel inclusion-exclusion games are utilised
as a bridge between semantical and syntactical fragments. These games
are not only the model-checking games of ESO-sentences, but they can
also be adapted for fragments of ESO and give rise to the definition of
a single team-based atom that captures the union-closed fragment.
Further we answer an open question due to Rönnholm regarding the
connection between inclusion logic of restricted arity and a fragment of
greatest fixed-point logic.
Finally, we study variants of logics with dependency concepts, which can
distinguish elements only up to a given equivalence. We juxtapose these
new logics with equivalent fragments of ESO and analyse their expressive
powers.
Es laden ein: die Dozentinnen und Dozenten der Informatik
+**********************************************************************
*
*
* Einladung
*
*
*
* Informatik-Oberseminar
*
*
*
+**********************************************************************
Zeit: Montag, 02. Dezember 2019, 13:30 Uhr
Ort: UMIC Research Centre, room 025, Mies-van-der-Rohe Strasse 15
Referent: Stefan Breuers, MSc., Lehrstuhl Informatik 8 (Computer Vision)
Thema: Multi-object Tracking and Person Analysis from Mobile Robot Platforms
Abstract:
Detecting and tracking the persons in a scene has been an active field of
research in the area of computer vision for decades. One of the main
applications are mobile platforms, such as autonomous cars or service
robots.
In this talk, we are going to analyze different tracking approaches working
on the image domain and have a deeper look at their errors, before we move
forward to multi-object tracking from robots operating in the 3D world.
We present our detection-tracking framework for systematic tracking
evaluation and our insights on experiments in challenging environments. On
top of that detection-tracking pipeline, we then run person analysis
modules and look at the benefits of such an integration for an improved
person-robot interaction. By utilizing the temporal information in the form
of the person track identities, we enable a robust and efficient operation
of expensive deep learning analysis methods on low-cost platforms.
We conclude the talk by presenting an explorative work on the integration
of person re-identification and multi-object tracking to overcome common
compromises of classical tracking approaches, taking a step towards an
end-to-end, image-to-track paradigm.
Es laden ein: die Dozentinnen und Dozenten der Informatik
+**********************************************************************
*
*
* Einladung
*
*
*
* Informatik-Oberseminar
*
*
*
+**********************************************************************
Zeit: Freitag, 22.11.2019 um 13.00 Uhr
Ort: Informatikzentrum, E3, Raum 5053.2 (ggü. AH 6)
Referent: Marius Alwin Shekow, M. Sc.
Informatik 5
Thema: Syncpal: A Simple and Iterative Reconciliation Algorithm for File
Synchronization
Abstract:
To facilitate collaboration scenarios and data management across multiple
devices, knowledge workers and individuals use file synchronizers. These
tools replicate the devices local file system to a cloud storage. As Marius
Shekow discusses in his doctoral viva, bugs in these file synchronizers
force users to detect and fix synchronization errors manually, resulting in
cost-intensive iterations in cooperation processes, which should be avoided.
He presents the three core challenges in file synchronization which he
examined in his doctoral thesis. You will learn about (a) heterogeneity of
file systems and alternatives to handle their incompatibilities, (b)
conflicting operations and options for conflict resolution, and (c)
inferring a valid operation propagation order to optimize the
synchronization efficiency and correctness. The speaker presents how these
challenges are solved by example with the iterative Syncpal algorithm he
developed in his thesis. The evaluation of its implementation shows that
Syncpal improves the handling of file system heterogeneity and synchronizes
changes from long offline periods correctly.
Es laden ein: die Dozentinnen und Dozenten der Informatik