Dissertations

in Progress

Completed

Viktoria Pammer

Automatic Support for Ontology Evaluation - Review of Entailed Statements and Assertional Effects for OWL Ontologies
The typical goal of representing knowledge formally is to make knowledge accessible to machines.
Hence, knowledge representation formalisms are chosen to accommodate well to the strengths of machines, often at the expense of being less easy to understand for humans.
At the same time, formal representations are most often created or maintained by humans. This discrepancy provides the motivation for the central idea of this thesis, namely to
facilitate evaluating the conceptual correctness of ontologies by means of reasoning services. For instance, when a logical axiom is added to an ontology, this addition may lead to
logically consistent but conceptually wrong inferences. In the context of this thesis, specifically description logics with a view on those underlying the W3C recommended web
ontology language OWL are considered as knowledge representation language in which ontologies are described. First, it has been investigated both by analysis and in practice how the review
of automatically inferred statements can serve the purpose of evaluating an existing ontology. In a second venture, the usage of instance data to give ontology engineers concrete examples
of the implications of general statements, i.e. terminological and role axioms in description logic languages, has been studied. Indeed, the effects of terminological and role axioms on
instance data have not been studied in literature so far, although from a mathematical point of view the question of conservative extensions, i.e. the logical difference between two sets of
terminological and role axioms, is very close. Thus it was necessary to provide a formal definition for effects that capture the intension of knowledge lost or gained about data, as well as
find and prove a decision procedure. Both research ventures have led to concrete implementations of ontology evaluation functionalities within MoKi, a wiki-based ontology engineering environment.
As a consequence, MoKi is now, to the best of the author's knowledge, the only state-of-the art ontology engineering environment that explicitly supports conceptual ontology evaluation rather
than ontology debugging in the sense of finding logical contradictions.

Andreas Rath

User Interaction Context - Studying and Enhancing Automatic User Task Detection on the Computer Desktop via an Ontology-based User Interaction Context Model

User context detection has recently gained momentum, thanks to the interest shown by several established research areas, such as information retrieval, personal information management, technology enhanced learning as well as task and process management. In the information retrieval area user context is exploited for personalizing search, in personal information management for relating tasks, processes, persons, documents and projects, in task and process mining for discovering task and process flows as well as in the fields of technology enhanced learning and knowledge work support for generating rich user profiles in order to provide appropriate learning and work material. Automatic task detection is one important challenge in the area of user context detection because once the user task is known it is possible to support her better. Automatic task detection on the computer desktop is classically seen as a machine learning problem, but has only been explored with text-based and switching sequence based user context features. Furthermore no public standard datasets from laboratory or real-world experiments for investigating task detection are available.

The goal of this dissertation research is to automatically detect the task of a user on her computer desktop in order to enable task-specific support which is especially important for knowledge workers facing complex tasks and information overload in today's knowledge economy. More specifically, this research strives to study and enhance automatic task detection on the computer desktop via an ontology-based user interaction context model (UICO). The user interaction context is a subset of the user context and is defined as
"all interactions of the user with resources, applications and the operating system on the computer desktop". The proposed ontology including the automatically observed user interaction context data is utilized for the feature engineering for automatic task detection, more specifically automatic task classification.

This thesis introduces a novel task detection approach referred to as the "ontology-based task detection approach" and evaluates it on three independent datasets
containing over 500 tasks from over 40 users of two different domains. These datasets are from three laboratory user experiments designed and performed as part of this research effort.
The most important insights gained from these evaluations are:
(i) combinations of features engineered from the UICO almost always outperform feature combinations suggested by existing task detection approaches,

Barbara Kump

Evaluating the Domain Model of Adaptive Work-Integrated Learning Systems
Adaptive learning systems have been suggested as highly promising means to support work-integrated learning. Adaptivity in a learning system is the result of a complex interaction between a domain model, a learner model and an instructional model. Existing approaches for evaluating adaptive learning systems take into account the learner model and the instructional model; the domain model, however, is mostly neglected. In my thesis, I conceptualize required properties of the domain model of an adaptive work-integrated learning system and derive four research questions: (Q1) Is the set of tasks exhaustive and conceptually correct?, (Q2) Is the set of skills exhaustive and conceptually correct?, (Q3) Is the requires relation conceptually correct (content validity)?, and (Q4) Are inferences made from the requires relation correct (application validity)? Then, I suggest a variety of methods for data collection and data analysis for answering these research questions. Each of the methods was applied in one or several out of 6 case and method studies in 4 different domains. Case Study 1 was a rigorous validation study in the domain of statistical data analysis (student domain) focusing on Q1, Q2 and Q3. In Case Study 2, Q1 and Q2 were looked at for the real work domain of aircraft simulation. Method Study 1 and 2 deal with the reliability and validity of self-assessment and peer-assessment. In Case Study 3, Q4 was tested in two real work domains, aircraft simulation and innovation management. In Case Study 4, a validation session was carried out in the innovation management domain to answer Q1 and Q2. Based on the findings from the studies, methodological considerations for each method are extensively discussed. The dissertation shows how the techniques can be used for validating the domain model of an adaptive work-integrated learning system, how they can be combined, and what conclusions can be drawn from the outcomes.

Peter Scheir

Associative Retrieval for the Semantic Web - A network-based search approach in consideration of semantic and content-based similarity

While in the current Web information is largely gathered by humans, the Semantic Web aims at having this task performed by computer programs. To achieve this goal information on the Web is enriched with machine-interpretable data that allows for identifying the meaning (semantics) of information on the Web. Approaches to search on the Semantic Web build upon this semantic metadata and integrate it as additional information into the process of search to increase retrieval effectiveness. While assigning semantic metadata to resources on the Web is seen as a central aspect for realizing the vision of the Semantic Web, only a small fraction of resources on the current Web is annotated with such metadata.

The work at hand contributes to the research field of search on the Semantic Web - currently being an early stage - by developing and evaluating an approach to information retrieval for the Semantic Web. In addition to exact search based on semantic metadata, methods stemming from associative information retrieval are integrated into the developed retrieval model. Associative information retrieval is a search paradigm which aims at identifying additional relevant information starting from information already known to be relevant and at increasing the performance of a retrieval system by this means. In the present work associations are based on semantic similarity between concepts from an ontology and on content-based similarity between resources. Besides an overall increase of retrieval effectiveness, the developed approach addresses situations in which few resources are annotated with semantic metadata, aiming at improving the results of a retrieval system that is based on the developed model. The model is implemented into a system for search on the Semantic Desktop and evaluated based on this implementation.

(in German language)

Johannes Farmer

AD-HOC - A Methodology for the Development of Virtual Work-and-Learning Environments

Organizations rely heavily on their knowledge workers to learn efficiently and to apply knowledge effectively within their work processes. Hence virtual work-and-learning environments are needed that support newcomers and experienced workers in equal measure and provide task-specific learning support ad hoc. However, by now contextualized learning support is costly, because it has to be tailored and introduced as a new system from scratch.

This problem has been solved with this thesis: an efficient design and implementation process for virtual work-and-learning environments - the so called AD-HOC Methodology - has been developed. With the AD-HOC Methodology learning needs are related to learning patterns. These patterns provide - similar to the object-oriented design patterns - predesigned learning support solutions for recurring learning problems. Then, the learning support solutions are implemented using generic functions. These functions map design requirements to communication, cooperation and information tools provided by the organization's information and communication (ICT)
infrastructure, in which the virtual environment will be embedded. Finally, the AD-HOC Methodology was applied in three case studies and tested successfully in different domains.

In summary, this doctoral thesis provides a powerful framework - the AD-HOC Methodology - for the efficient design and implementation of virtual work-and-learning environments for the first time.

Keywords: Knowledge Management, E-Learning, CSCL, Knowledge Work, Learning
Pattern Language, Generic Functions, Productivity, System Design.

(in German language)

Tobias Ley

Organizational Competency Management - A Competence Performance Approach

This work is concerned with a new approach in organizational Competency Management.

The goal is to develop a method that is practically feasible for organizational settings, is firmly based in psychological conceptions of human competence and performance in the workplace, and employs a degree of mathematical formalization that improves possibilities for establishing the validity of the implementation. The approach is developed on the basis of the Competence Performance approach (Korossy, 1997, 1999) which is based on the fundamental idea of establishing prerequisite relations on the set of competencies and performances, so as to allow efficient assessment. The structures that can be derived from these relations can also be interpreted as formalizing learning paths on the competence and performance level. The approach is applied in one case study in the automotive industry and two empirical investigations in a dynamic research based setting. In the empirical investigations, documents are used as performance outcomes. The repertory grid technique is used as an interview method. The competence performance matrix is introduced as a means to derive competence performance structures. For evaluation purposes, quality criteria which measure reliability and validity of the resulting models are introduced. A scenario for work integrated competency assessment and development shows the practical relevance of the approach for workplace learning.

Summary:

Full text:

Tobias Ley: Organizational Competency Management – A Competence Performance Approach (ISBN 3-8322-5051-4)

is published at Shaker Verlag

Thomas Dietinger

Aspects of e-Learning Environments

Tremendous advances in computer technology and the evolution of the Internet have led to new approaches in learning and training which are summarized under the term e-Learning. This thesis will explain what can be expected from e-Learning and although especially focused on the technological basis, will examine all kinds of requirements for e-Learning environments: pedagogical, functional and non functional requirements. A special chapter will be dedicated to all relevant standards in the field of eLearning due to its importance for increasing interoperability, cutting costs and gaining acceptance.
An example for a state of the art e-Learning system, which has been designed by and developed under the guidance of the author, will be given, successfully completed projects based on this environment will be presented to show possible operational areas, and ideas for further developments will be sketched.

Markus Strohmaier

B-KIDE: A Framework and a Tool for Business Process Oriented Knowledge Infrastructure Development

The need for an effective management of knowledge is gaining increasing recognition in today’s economy. To acknowledge this fact, new promising and powerful technologies have emerged from industrial and academic research. With these innovations maturing, organizations are more and more willing to adapt such new knowledge management technologies to improve their knowledge intensive businesses. However, the successful application in given business contexts is a complex, multidimensional challenge and a current research topic. Therefore, this PhD thesis addresses this challenge and introduces a framework for the development of business process-supportive, technological knowledge infrastructures. While business processes represent the organizational setting for the application of knowledge management technologies, knowledge infrastructures represent a concept that can enable knowledge management in organizations. The B-KIDE Framework introduced in this work provides support for the development of knowledge infrastructures that comprise innovative knowledge management functionality and are visibly supportive of an organization’s business processes. The developed B-KIDE Tool eases the application of the B-KIDE Framework for knowledge infrastructure developers. Three empirical studies that were conducted with industrial partners from heterogeneous industry sectors corroborate the relevance and viability of the introduced concepts.

Dolly Kandpal

Augmenting Knowledge-Based Systems with Dynamic Personalisation Concepts

In the knowledge-intensive society of today, the emphasis is on delivering the right kind and quantity of just-in-time knowledge to the knowledge workers. Personalization, i.e. the adaptation of a system according to a user's needs, is gaining increasing relevance as an enabler for delivering such relevant knowledge. Personalization can be of type static or dynamic. Static personalization takes place at compile or design time, whereas dynamic personalization takes place at run-time. Static personalization is limited in that it defines a personalization at a certain point in time reflecting only the current interest and knowledge background of a knowledge worker. The limitations are obvious: dynamic changes of the interest and knowledge cannot be covered by the existing concepts. This is why concepts for dynamic personalization are becoming increasingly important. The main idea of dynamic personalization is to provide support for knowledge workers, allowing them to define personalization features which continuously adapt themselves to the needs, interest, context and environment in which the knowledge workers operate. Currently, although a number of scientific approaches deal with static personalization in different domains, very little work has been done to address dynamic personalization at a conceptual level. Also, an extensive study of existing knowledge-intensive systems in various domains reveals that many of them already offer features that support dynamic personalization. But from a scientific point of view, no sound conceptualization of the main aspects of dynamic personalization exists.

The objective of this work is to treat dynamic personalization at a conceptual level. This is achieved by establishing a common understanding and definition of dynamic personalization, and by introducing a framework for dynamic personalization of knowledge-intensive systems based on the current user context. The user context comprises not only of the user preferences and profile information, but also environmental aspects like time, location, access device and network characteristics. The framework comprises of an abstract meta-model for providing dynamic personalization, the domain-specific models, and the Mapping methodology. By using the methodology (a sequence of guidelines) called the Mapping methodology, the generic meta-model is used to derive dynamic personalization models for various knowledge-intensive domains, called domain-specific models in the context of this work. The dynamic personalization features of models and applications in various knowledge-intensive domains can then be modeled using the dynamic personalization features of the domain-specific model. The framework is held generic to enable its applicability to a number of knowledge-intensive domains where dynamic personalization aspects are required.

Herwig Rollett

Knowledge Management Technologies in Context

Most of the knowledge management literature to date is written either from a business perspective, tending to ignore important technological considerations, or from a technology perspective, tending to ignore the larger organizational context. By contrast, this thesis focuses on the role of technologies in knowledge management without losing sight of organizational issues, firmly placing the discussion of technologies in a business context.
Two different but complementary views of knowledge management are presented, one based on processes, the other one on interactions. Lessons learned from past knowledge management projects are collected, summarized, categorized, and used to highlight the most pressing issues in knowledge management. Seven knowledge management processes and their relationships are discussed in detail: Planning, creating, integrating, organizing, transferring, maintaining, and assessing knowledge. Technologies with the potential to support knowledge management are split into eleven groups, and their relevance to knowledge management is systematically examined by assessing the contribution of each group of technologies to each of the knowledge management processes. Many open issues and opportunities for future research and development are identified, both organizational and technological. A very comprehensive knowledge management bibliography concludes the thesis.

Extract from the Book Knowledge Management 272,72 kB

Order from Springer

Diploma Thesis

Completed

Barbara Kump

Technology-Enhanced Workplace Learning: Modelling Competence and Performance for the Case of Requirements Engineering

From a psychological point of view, two models are required for the construction of an adaptive learning environment for informal, self-directed work-integrated learning (e.g., APOSDLE): the Knowledge Base, a model of the structured expert knowledge of a domain, and the S tudent Model, a model of the competencies of the learners. Based on these two components, individual learning paths can be optimised and appropriate learning resources can be selected.

In the course of this work, a modelling method should be developed in order to model the Knowledge Base and the Student Model for an adaptive learning environment. The theoretical framework chosen was the Competence Performance Approach, a mathematical-psychological theory on human learning and performance. One of the decisive advantages of Competence Performance theory is that it allows for integrating the Knowledge Base and the Student Model of an adaptive learning environment into one single model.

The developed modelling method was examined for the learning domain Requirements Engineering and the resulting models were evaluated in a case study. Both, document analyses and systematic expert interviews were conducted in order to acquire domain knowledge.

Karl Köberl

Acquisition of User Information with AJAX

As a result of the huge amount of information in the recent years it is hard for a user to find important information. A solution is to use personalized systems in an intranet system or an Internet application. For a personalized system a user model is needed. To get a model, some information of a user must be captured. Implicit feedback is retrieved to get long term interests of a user. With this information a user model is generated.

First in this thesis some information about the topic interest indicators are gathered. Then a own taxonomy of evaluable implicit interest indicators has been developed. In this taxonomy are nine indicators which describe the user’s interests. The captured indicators provide a basis for a generic user model. In this thesis a ”history-based-model“ is used. In the practical part of the thesis a web application to gather information of the user is implemented. This system can track user behaviour and build a user model. For the application the new technology AJAX (Asynchronous JavaScript and XML) is used. So a part of this thesis gives a detailed introduction in AJAX.

(in German)

Thomas Moser

Steuerungsinstrumente und Maßnahmen für Wissensrisiken

As a result of rising competition in most branches, new laws and regulations the need for effective risk management increases. Effective risk management requires critical risks identification and treatment. In this paper the author wants to determine how far the task of risk response planning can be supported.

For this purpose, the author combines the key risks for companies with knowledge intensive processes and business operations to a new definition. The definition distinguishes between risks based on knowledge gaps and risks which endanger the resource knowledge. Based on this definition for the term knowledge risk, the author formulates the Risk Response Planning Model and implements it in a software prototype for an easier application.

Essential parts of the model are a catalog on generic risks and a catalog on adequate risk responses. These two catalogs are linked together with a logic, which is developed by the author. If a user instances a new specific risk from a generic one, the model will propose adequate generic risk responses. Thus the model and the prototype support the user by the planning of adequate risk responses for knowledge risks.

Elisabeth Hackl

The Knowledge Map as a Tool to Support the Reuse of Project Related Information

Knowledge Maps are generally seen as tools that increase efficiency within organisations by facilitating the location of a company's knowledge and therefore avoid redundancies. This thesis aims to answer the question, how the Know-Center's Knowledge Map has to be designed to support the reuse of project related solutions and knowledge.

Based on related research on Knowledge Maps a concept is developed showing how to support the target group's information needs regarding project related information. In addition, the tool will be evaluated in view of the research question.

Manuel Fischer

A Contribution to Affinity Measures in Knowledge Management Systems

Starting point of this thesis is experience that has shown that additional value is created in knowledge-management systems, if the relationship between knowledge carriers is made explicit to users. Knowledge carriers can be users as well as documents, in this context. Based on this, the goal of this work is to identify and measeure correlations within a documentbase, between users among themselves and between documents and users.
Concepts for this functionality can be found in the field of people locator and recommender systems. The application environment is PADDLE - "Personal Adaptable Digital Library Environment" - developed with participation of the Know-Center. PADDLE provides the document base and will be extended by the mentioned functionality.
As the foundation similarity measures on the basis of meta-data of different types, like thematic, temporal or geographic descriptions, are developed to identify similarity between knowledge carriers. The primary focus is on thematic meta-data, describing the content of a knowledge carrier with a few keywords. These concepts are implemented prototypically and integrated into PADDLE. Evaluation is carried out to compare results of the measures with different sets of parameters with the subjective jedgement of "similarity" within the documentbase of PADDLE.

Herbert Pacnik

Blended Learning - Integration of eLearning with Traditional Teaching Methods

eLearning can be introduced from the scratch within corporate environments only in rare cases. Most often traditional training concepts are well established and need to be taken into account when developing an integrated concept for eLearning. Much literature is available on how to build eLearning right from the beginning, but hardly anyone addresses the process of seamlessly integrating eLearning into existing traditional training structures. The purpose of this theses is to describe such an integration process together with a real world example. It also shows how to deal with such integration using a a well known eLearning system.

Eduard Stoisser

Information Technologies for Knowledge Networks - Tools in the Context of Knowledge Management

Networks for knowledge creation, knowledge use and knowledge transfer are well known since existence of all forms of human organization. A knowledge network is defined as the deliberate interlinking of knowledge carriers leading to cooperation and common use of knowledge. The role of information technology in this field is currently the topic of fierce debates by researchers and practitioners alike. While supporting technology is not a condition for the creation and persistence of knowledge networks, it helps to improve the general context in a lot of areas. This work discusses temporal relevance, use and suitability of different information technologies for support of knowledge networks. The main emphasis lies is the examination of integrated solutions within a knowledge management framework. By analysing applied contexts, ways of deploying information technologies for knowledge networks with respect to the particular needs are being derived.

Michael Hausenblas

Semantic Representation and Query of Legal Norms in the Area of Academic Law

The motivation for this work is driven by solving the problem of retrieving relevant legal fragments in the realm of the austrian academic law starting from the statement of affairs. This can be seen as the legal process of subsumption, i.e. to find a passage in a law text that regulates a matter of fact.
Starting with this goal, the idea of creating a high level language for the Semantic Web got more and more important. This direction of impact can be justified with the increasingly spread of Semantic Web vocabularies like Topic Maps. Still the human factor is the ultimate limitation in the utilization of these techniques. Quite frankly, Semantic Web languages nowadays are, what the Assembler language is to third and fourth generation languages like LISP and Java. Using high level languages can enormously increase the creation and maintenance of knowledge bases and ontologies.
Last but not least this work proposes a new way of knowledge representation. Based on Heraklits Panta Rhei, an ontology representation founded on the description of so called threads is given.
This paper and its realisation is assigned to the area of applied AI and should be understand as a contribution to the evolving Semantic Web with a strong interest on the benefits of the practical results.

Leander Härter

Strategies and Tools for Structuring and Managing Heterogenous Datasets

Information created by several sources and target groups is complicated to organise. Different requirements need to be considered in order to accurately draw up a process for the target groups. This Master Thesis is describing strategies and tools which serve the administration of the heterogeneous data set, and should improve the work with this system. Three different areas displaying the information assimilation cycle are being looked at: The area "Publish" deals with the process of how information is stored in a database. In the area "Search and Navigate" aids and tools can be found which help to arrive at the requested documents. Finally, the area "Maintenance" describes ways on how the existing tools can be made suitable to deal with the requirements of a growing system.

Peter Scheir

Supporting Customer Relationship Management through Knowledge Management - Concepts, Technologies and a Prototype

This thesis shows how knowledge management can be applied for the successful realization of customer relationship management activities. First, this thesis introduces the topics 'knowledge management' and 'customer relationship management'. Next, aspects of both are combined into knowledge management in customer relationship management or 'customer knowledge management', for short. A concept showing how to support various persons in acquiring customer knowledge is presented. Based on those theoretical insights an information system to support employees concerning customer knowledge is implemented. Additionally, convenient technologies for this implementation are evaluated.