A-2004-2 Markku Turunen, Jaspis
– A Spoken Dialogue Architecture and its Applications.
February 2004.
Abstract.
Speech can be an efficient and natural way for communication between
humans and computers. Many practical applications have been
constructed,
but the full potential of speech applications has not been utilized.
In
addition to technological shortcomings, the development of speech
applications lacks suitable techniques, methodology and development
tools. For example, mobile and multilingual communication needs
flexible
and adaptive interaction methods which take into account the needs of
different users and different environments. This dissertation
addresses
the following question: what kind of a system architecture do
advanced
speech applications require? The following challenges are
specifically
addressed: How could the system architecture support advanced
interaction techniques? How could application development be
supported
by suitable models, methodology and tools?
This dissertation introduces the Jaspis speech application
architecture
that has been designed to support adaptive and flexible
human-computer
interaction techniques. This work also presents several applications
constructed on the Jaspis architecture. Two multilingual e-mail
applications, and two timetable applications are presented to provide
concrete examples. The challenges of pervasive speech applications
are
introduced in the context of an ubiquitous computing application. The
findings from the use of the Jaspis-based applications are
reported.
Several Jaspis-based interaction models and tools are introduced to
facilitate the development of practical applications. The focus is on
human-computer interaction issues, and solutions are presented for
tasks
such as error management, Wizard of Oz experiments and corpora
collection. Finally, an enhanced version of the Jaspis architecture
is
presented. The Jaspis2 architecture focuses on distributed and
concurrent spoken dialogues, which are needed in pervasive
applications,
and in general to provide more natural interaction models. The main
contribution of this work, the Jaspis architecture, will be released
simultaneously with this dissertation as an open source software for
speech application researchers and developers.
Ph.D. Dissertation.
A-2004-2 has appeared electronically as
Acta Electronica Universitatis Tamperensis, vol. 325.
A-2004-3 Poika Isokoski, Manual Text Input:
Experiments, Models, and Systems.
April 2004.
Abstract.
Despite the emergence of speech controlled computers and direct
manipulation that both have diminished the need to operate computers
with
textual commands, manual text entry remains one of the dominant forms
of
human-computer interaction. This is because textual communication is
one
of the main reasons for using computers.
Mobile and pervasive computing have been popular research areas
recently.
Thus, these issues have a major part in the thesis at hand. Most of
the text entry methods that are discussed are for mobile computers.
One
of the three main contributions of the work is an architecture for a
middleware system intended to support personalized text entry in an
environment permeated with mobile and non-mobile computers.
The two other main contributions in this thesis are experimental work
on text entry methods and models of user performance in text entry
tasks.
The text entry methods tested in experiments were the minimal device
independent text entry method (MDITIM), two methods for entering
numbers
using a touchpad, Quikwriting in a multi-device environment, and a
menu-augmented soft-keyboard. MDITIM was found to be relatively
device independent, but not very efficient. The numeric entry
experiment
showed that the clock metaphor works with a touchpad, but with a high
error rate. An improved "hybrid" system exhibited a lower error rate.
Quikwriting was tested to evaluate the claims on its performance made
in
the original publication and to see if it works with input devices
other
than the stylus. The perfomance claims were found to be exaggerated,
but
Quikwriting worked well with the three tested input devices (stylus,
game
controller, and a keyboard). The menu augmented soft keyboard was
compared
to a traditional QWERTY soft keyboard to verify modeling results that
show
significant performance advantages. No performance advantage was
observed
during the 20 session experiment. However, extrapolations of the
learning
curves cross suggesting that with enough practice the users might be
able
to write faster with the menu augmented keyboard.
The results of the modeling part are two-fold. First, the explanatory
power of a simple model for unistroke writing time was measured. The
model accounted for about 70% of the variation when applied carefully,
and
about 60% on first exposure. This sets the level of accuracy that
more
complex models must achieve in order to be useful. Second, a model
that
combines two previously known models for text entry rate development
was
constructed. This model improves the accuracy of text entry rate
predictions between measured early learning curve and the theoretical
upper limit.
Ph.D. Dissertation.
A-2004-3 has appeared electronically as
Acta Electronica Universitatis Tamperensis, vol. 340.
A-2004-4 Kari Kilpinen, Reflective teacher using the
computer-network in teaching; how the psycho-epistemological learning styles help
to better design learning environments. May 2004.
Ph.D. Dissertation.
A-2004-4 has appeared electronically in
Acta Electronica Universitatis Tamperensis, vol. 365.
A-2004-5 Timo Niemi, Marko Junkkari, and Kalervo Järvelin,
Query language approach for next generation information systems. April 2004.
Abstract.
In developing next generation information systems (NGISs) three kinds of integration
is necessary: the integration of data-oriented, behavioral and deductive aspects of
application domains, the seamless integration of the manipulation of the essential
relationship types (is-a relationship, part-of relationship, association) and the
integration of the manipulation of the extensional and intensional levels of data.
Further, NGISs have to support the manipulation of text documents which can be associated
with entities or relationships among them in an application domain. Information in
text documents can also be used as sources for deduction. In this paper we address
these topics. We introduce a set of primitives which support required integration
in a natural and flexible way. Based on these primitives we have developed a
concept-oriented query language for NGISs. Its user can make expressive and compact
queries by combining available concepts and concept structures equipped with variables.
Our sample queries demonstrate that, in addition to conventional extensional queries,
our query language also supports intensional and combined extensional – intensional
queries which are beyond of the capabilities of contemporary query languages.
Keywords: Next Generation Information Systems (NGISs), Query languages,
XML documents, Extensional and intensional levels of data, Integration of data.
A-2004-6 Jaakko Riihimaa, Taxonomy of
information and communication technology innovations adopted by small and
medium sized enterprises. May 2004.
Ph.D. Dissertation.
A-2004-6 has appeared electronically in
Acta Electronica Universitatis Tamperensis, vol. 366.
A-2004-7 Timo Poranen, Approximation Algorithms for the Topological Invariants
of Graphs. September 2004.
Abstract.
Topological graph theory studies the embeddings of graphs on various
surfaces and the properties of these embeddings. Topological properties of
graphs have many applications in the fields of graph drawing, user interface
design, circuit design, and resource location optimization.
This thesis studies approximation methods for the following
NP-complete optimization problems: maximum planar subgraph, maximum outerplanar subgraph,
and thickness of a graph. We also study the outerthickness problem
which complexity status is not known.
We compare the solution quality and computation times of a simulated
annealing algorithm and
several algorithms based on triangular cactus heuristic, including
other heuristics taken from the literature, to approximately solve
these problems.
Triangular cactus heuristic was the first non-trivial approximation
algorithm for the
maximum planar and outerplanar subgraph problems. We give a modified
version of the triangular cactus
heuristic that has at least equal performance ratio and asymptotic
running time as the linear time version of the original algorithm.
A large number of experiments show that the new algorithm achieves
better approximations than the earlier methods.
We give two new theoretical results for the thickness and
outerthickness of a graph.
We prove a new upper bound for the thickness of complete tripartite
graphs, and lower
and upper bounds in the terms of the minimum and maximum degree of a
graph for outerthickness.
Also, the simulated annealing algorithm given in this work solves
partially an open problem related to
the thickness of the complete bipartite graphs. Our experiments show
that the general formula also holds for some previously unknown cases.
Ph.D. Dissertation.
A-2004-7 has appeared electronically in
Acta Electronica Universitatis Tamperensis, vol. 391.
A-2004-8 Marko Junkkari,
Modeling and manipulation of composed objects in NGIS. December 2004.
Abstract.
In Next Generation Information Systems (NGISs) object-orientation and deductive
aspects are integrated to databases. This means that any modeling methods
intended for NGISs need integrate data-oriented, behavioral and deductive
aspects of application domains. RDOOM (Relational Deductive Object-Oriented
Modeling) is a modeling method in which these aspects are seamlessly integrated
with each other. However, until now RDOOM has not been complete to model such
objects which contain other objects (called composed objects). In other words
among the principal data modeling relationships RDOOM has only dealt with the
association and the is-a relationships but not the part-of relationship which is
the third principal modeling relationship. Here, we extend the modeling of NGISs
by the manipulation of the part-of relationship. This work involves integration
of is-a and part-of hierarchies, deduction based on part-of structures, general
methods aimed at manipulate structurally different composed objects, and
manipulating unknown aspects in part-of hierarchies. In this paper, we present a
comprehensive set of data modeling primitives on which conceptual derivation can
be based.
Keywords: NGIS, data integration, formal specification,
conceptual derivation, data model, deduction, composed object, part-of
relationship, query languages, real word applications, inheritance, aggregation
1992-2001
2002
2003
2005
2006
2007
2008
2009
To the upper level