Research projects

Contact person: Kwabena Odame Akomeah (opens your email program)

A Social Media Companion Safeguarding and Educating Students

This project aims to develop a Social Media User's Virtual Companion that educates and supports teenage school students facing the treats of social media such as discrimination and biases also escalating to hate speech, bullying, fake news and other toxic content that can strongly affect the real world.

The Companion will raise awareness of potential threats in social media while still providing a satisfactory experience through the use of novel gamification strategies and educative information selection algorithms. Using hand defined gamification strategies based on the concept of interactive counter-narrative together with a learnt model of diffusion of biased, hateful and general toxic content on the social media, the companion will bootstrap learning of strategies for interaction and information selection.

These strategies will be refined and personalised for each user and his social niche with the dual aims of (A) improving and creating healthy social relationships between the user, his peers and the targets of bias, as well as(B) increasing their understanding of the social effects of toxic content in social media and the user role in its propagation.

Experimentation will use data mined from real social media and re-enact them for testing in restricted and controllable conditions (eg. schools, classes).

Team:

• Prof. Dr Udo Kruschwitz

Co-operation partners:

University of Applied Sciences Ruhr West (external link, opens in a new window) (Bottrop, Germany)

L'Istituto per le Tecnologie Didattiche (external link, opens in a new window) - CNR-ITD (Palermo, Italy)

University of Essex (external link, opens in a new window) (Colchester, UK)

Universitat Pompeu Fabra (external link, opens in a new window) (Barcelona, Spain)

Rhine-Ruhr Institute (RIAS) (external link, opens in a new window), (Duisburg, Germany)

University of Milano-Bicocca (external link, opens in a new window) (Milano, Italy)

Funding:

Volkswagen Foundation (external link, opens in a new window)

Contact person:

Prof. Dr Udo Kruschwitz

FoodChoice - Modelling & Changing Online Food Choices through Recommenders

Studying how people upload and browse online recipes has recently become an active field of research. While there is a growing body of work investigating which types of food are consumed and how this relates to real-world health related issues, such as e.g. diabetes or obesity, little research focus has yet been devoted to understanding how people make their food choices online, how they are influenced by certain contextual factors, how this behaviour can be modelled or how this behaviour can be changed. With respect to the last problem, recommender systems - such as those used by online stores to suggest products customers might like- are often touted as part of the solution. However, very little literature exists describing to which extent food preferences can be changed or the different effects different algorithmic approaches can have. In response we initiated a cross-border research project FoodChoice, which aims not only to provide a better understanding regarding how certain factors, such as temporality, user geography and social relations between users influence the food choices people make online, but to moreover investigate how this behaviour can be modelled and changed through novel health-aware recommender algorithms.

Team:

• PD Dr David Elsweiler
• Dr Christoph Trattner (University of Bergen)
• Prof. Dr Bernd Ludwig

Where to Go and What to Do: Towards Understanding Task-Based Information Behavior at Transitional Spaces

In public transitional spaces, such as airports, users are faced with diverse challenges regarding information interaction and use. These challenges arise due to the scheduled and/or location-dependent procedures users are required to perform. Understanding what these users need or desire in the context of such spaces, what information is on offer, both online and in situ, and how these aspects interrelate is important to facilitate the design of systems that are accepted by the users concerned. However, very little is known about human information behaviour (HIB) in public transitional spaces. As a starting point to understand how behavior in such spaces relates to or differs from information behavior in other contexts, holistically, I will create an explanatory model of airport information behaviour by conducting an exploratory grounded theory based field study and relating my findings to those of existing models.
(from: Melanie A. Kilian. 2019. Where to Go and What to Do: Towards Understanding Task-Based Information Behavior at Transitional Spaces. In 2019 Conference on Human Information Interaction and Retrieval (CHIIR '19), March 10-14, 2019, Glasgow, Scotland Uk. ACM, New York, NY, USA, 4 pages. doi.org/10.1145/3295750.3298972 (external link, opens in a new window))

Team:

• Melanie A. Kilian, M.A.
• PD Dr David Elsweiler
• Prof. Dr Bernd Ludwig

Contact person: Melanie A. Kilian, M.A. (opens your email program)

Corroborate: Understanding and supporting the information credibility judgments people make online

Making credibility judgements in information environments, such as the web is challenging because, in contrast to traditional print and broadcast media, often no quality control mechanism exists. This places increased emphasis on user information literacy skills, including their ability to evaluate information critically. These are skills which people, regardless of education level, tend to overestimate (Gross 2012 (external link, opens in a new window)). Furthermore, recent IR research has emphasised that user actions while seeking information, as well as the perception of information found is biased in countless ways. Although credibility judgements for web pages and SERP snippets have been well studied, we still know relatively little about how such evaluations are made and how they may be affected by user biases. This project aims to broaden our understanding of how people judge credibility, how successful they are, and what determines this. We further aim to design systems to support the critical evaluation of information informed by the improved better understanding the processes and biases involved.

Team:

• PD Dr David Elsweiler
• Dr David Losada (Universidade de Santiago de Compostela)
• Dr Markus Kattenbeck (Technical University of Vienna)

URWalking - University of Regensburg Walking

Pedestrian navigation with landmarks

Do you remember the last time you had to ask for directions? Were you told to turn right in 250 metres? Probably not, because directions are usually communicated with the help of conspicuous objects (landmarks).
As part of the UR-Walking project (external link, opens in a new window), this form of communication is also to be made available for mobile navigation systems. The labyrinthine campus of the University of Regensburg is particularly suitable for trialling landmark navigation. The topic is also very popular with students in seminars

Two participations at CeBIT (external link, opens in a new window) testify to the very successful progress of the URWalking project to date.

Flyer CeBIT 2012 (opens in a new window). (This PDF is not accessible)

Flyer CeBIT 2013 (opens in a new window). (This PDF is not accessible)

UR-Walking - PromoFilm 2012 (external link, opens in a new window)

UR-Walking - PromoFilm 2013 (external link, opens in a new window)

Project manager: Prof Dr Bernd Ludwig (opens your email program)

An Empathetic Conversational Agent for Weight Loss Treatment

The increasing rate of obesity in the German population calls for the development of new treatment options as a way to reduce costs on the public health care system and increase treatment success.

This project explores possibilities of automated empathetic conversation to support traditional obesity treatment. The constant availability of a conversational agent could solve problems related to weight loss as soon as they arise and help with impulse control. The non-judgemental nature of such a system as opposed to counsellors or peers has the potential to lower the user's inhibitions and increase honesty in conversations about sensitive topics.

In the long term, such a conversational agent could be a valuable tool in the treatment and prevention of obesity.

Team:

• Dr Selina Meyer
• PD Dr David Elsweiler
• Prof Dr Bernd Ludwig

Contact person:

Dr Selina Meyer (opens your email program)

Information Needs in Natural Language Dialogues

As conversational search becomes more pervasive, it becomes increasingly important to understand the users' underlying needs when they converse with such systems in diverse contexts. The aim of this project is to improve information need detection. In the long term, this will help improving conversation between user and system, because the system gains a better understanding of what the user wants when it is able to detect the user's information needs. For this purpose, we have chosen the food domain as area of interest.

Team:

• Dr Alexander Frummet
• PD Dr David Elsweiler

Contact person:

Dr Alexander Frummet

Understanding food culture with digital traces

The aim of our project is to use the traces of human behaviour with online recipes to explore the differences in food culture across the world. We are now focusing on the visual aspects of the images associated with recipes. Based on the Explicit Visual Features (EVF) and Deep Neural Network image embeddings (DNN), we plan to dig out how food culture is distributed and the aesthetics of preferred recipe images within and across cultures. Techniques related to computer science, information science and data science would be employed in our project. We have done a couple of experiments on three datasets which sourced from three large recipe platforms from China, Germany and the United States, which shows that using visual features as mentioned above, 1) we could train classifiers to differentiate the images from these three collections; 2) it is also possible for us to predict if a recipe is appreciated or not, although the accuracy is not so satisfied. 3) visual tastes of German and US are more similar to each other than those of the Chinese.
These findings suggested that there are indeed differences in the visual aspects of online recipes. In the future work, we will cooperate with anthropologists to figure out how the differences generated and how to explain them.

Team:

• Dr Qing Zhang (Information Science, UR)
• PD Dr David Elsweiler (Information Science, UR)
• Dr Christoph Trattner (Information Science and Media Studies, University of Bergen)

Contact person:

Dr Qing Zhang (external link, opens in a new window)

OPTAPEB - Optimising psychotherapy through agent-guided patient-centred emotion management

Phobias trigger panic attacks in millions of people and can severely restrict them in their everyday lives. Psychological treatment is usually carried out with the help of exposure sessions in which the anxiety trigger is dealt with in a controlled manner.

In the "OPTAPEB" (external link, opens in a new window) project, a system was developed that enables VR-supported exposure therapy with a virtual agent. The client's emotional reactions during the exposure were processed by a multimodal sensors system with subsequent data fusion. Therapeutic micro-interventions were derived from this, which were implemented by the virtual agent or the environment. This allowed the therapeutic treatment to be meaningfully extended and, for example, continued and deepened in the living room at home.

Network coordinator: University of Regensburg

Project partners:

• VTplus GmbH, Würzburg
• Fraunhofer Institute for Integrated Circuits (IIS), Erlangen
• Ambiotex GmbH, Tübingen
• New Textile Technologies GmbH, Balingen
• audEERING GmbH, Gilching
• ZTM Bad Kissingen GmbH, Bad Kissingen

Duration: 09/2017 - 08/2020

Funding: Federal Ministry of Education and Research (BMBF)

Calculation of routes taking into account general and individual user preferences

As part of the URWalking doctoral project by Dipl.-Inf. Manuel Ullmann, the aim was to present users of the navigation system with sensible alternatives to the shortest route. The behaviour of users with local knowledge when planning routes on campus was investigated.

This behaviour was then simulated using methods from the field of machine learning. This made it possible to suggest more natural routes, but also to imitate the individual behaviour of individual users and user groups.

ATiZu@Web - Design and implementation of a research support platform as part of the EU project "From Tradition to the Future"

The INTERREG project "From Tradition to the Future" (external link, opens in a new window), in which the Universities of Passau and ?eské Bud?jovice participated, aimed at "Researching the linguistic and literary heritage of South Bohemia" (see there).

As part of a subcontract for this project, a web application in Python was implemented using the "Django" framework, which facilitates data entry for research purposes, offers an end-user view of this data and also offers a wide range of search options, including map-based search options. In addition, a mobile application was developed that enables interested laypersons to access the research results and, for example, to compile routes to places of linguistic and literary interest.

Co-operation partners: Universities of Passau and Budweis

DIVIS - Digital indoor localisation using inertial sensors and landmarks

In the "DIVIS" research project, a positioning technology for pedestrian indoor navigation systems was developed in cooperation with NEXOE Applications GmbH from Regensburg, which can be used to navigate people safely to their destination. The positioning was not only carried out using conventional sensors integrated in the smartphone, but also in dialogue with the user using image and voice recognition.

Duration: 08/2016 - 10/2019

Funding: R&D programme "Information and Communication Technology" of the Free State of Bavaria (Freistaat Bayern)

URroadworks: Navigating in dynamic environments

In URroadworks, the scalability of the URWalking data model and the time required to make even complex changes to the model at short notice are evaluated in a realistic scenario. In the coming years, the central part of the Regensburg University campus will be redeveloped. The construction work will constantly lead to new closures of familiar routes, especially for people with disabilities.

As a countermeasure, people on blocked paths are made aware of the campus navigation system by means of diversion signs with QR codes. As soon as a person has scanned a QR code, their position is known and they can be guided to their destination by URWalking. In the background, the navigation data in URWalking's data model is regularly checked and updated so that URWalking's navigation instructions are always up to date.

The usage data will later be used to empirically validate theories about the information behaviour of people in public spaces: what information needs are created by the construction site? Which system responses are helpful in which situation and why? How do users with local knowledge behave towards the unusual route instructions - will they try to find their own shortcuts? How do people who do not know the area behave?

Project manager: Prof Dr Bernd Ludwig (opens your email program)

App-based support for changes in nutritional behaviour

Elinor Brondwine M.Sc. (external link, opens in a new window) was a ZD.B (external link, opens in a new window) doctoral scholarship holder whose research project focussed on nutrition in the context of healthcare. The aim of the project was to help patients to implement nutrition-related instructions from medical staff over a longer period of time.

Funding: Zentrum Digitalisierung.Bavaria Fellowship (external link, opens in a new window) (ZD.B Fellowship), BayWISS-Verbundkolleg Gesundheit (external link, opens in a new window)

URWalking sub-project: Supporting orientation in indoor areas - analysis of landmark-based map interfaces based on the gaze behaviour of users

As part of Dr. Christina Bauer's doctoral project, a sub-project of URWalking, different map displays that refer to conspicuous objects were evaluated based on the gaze behaviour of the users. A mobile eye tracker was used for this. The research focussed on the reception of the interfaces and the environment during navigation in the interior.

Re-finding tweets - analysing the personal information management practice of re-finding in the context of the social media platform Twitter

As part of his dissertation project, Florian Meier (external link, opens in a new window) focussed on re-finding, i.e. the retrieval of information that has already been found.

Refinding is an everyday problem in the administration of information that helps us to manage our daily lives (Personal Information Management, PIM). In addition to printed documents, addresses, e-mails, websites and search engine results, this also includes messages on social networks. It has been proven that social networks are not only used to maintain and cultivate social contacts, but are also a source of specific information (links), personal experiences or are used as question and answer systems.

Florian Meier specifically analysed the problem of information re-finding on Twitter. This includes aspects such as:

• What motivates users to re-find previously seen tweets?
• What methods do they use for this?
• How does Twitter support them in this task?

Finally, the dissertation project posed the question of how search user interfaces for social networks can be improved with regard to the need for information "re-finding".

As part of his dissertation project, Dipl.-Inf. (FH) Markus Fuchs M.A. focussed on feedback systems in university teaching:
Classgame is a playfully designed online learning platform that was developed for programming training and is intended to reduce cognitive load.

The system is still used in the Information Science Group's teaching today and is constantly being adapted.

URTalking - University of Regensburg Talking

"URTalking" was a research project of the Chair of Information Science at the University of Regensburg. As part of the seminar "Methods of natural language processing", Master's students developed a chatbot prototype under the guidance of apl. Prof. Dr Reischer, which was intended to support prospective students and students in their studies at the University of Regensburg.

Project manager: Prof. Dr Jürgen Reischer (opens your email program)

NADINE - Navigation in public transport

As part of the project "NADINE - Navigation in public transport with modular service architecture for integration into external applications (external link, opens in a new window)", the University of Regensburg worked on the implementation of a landmark navigation system. The project was carried out as part of the funding programme "From door to door - a mobility initiative for the public transport of the future" of the Federal Ministry of Economics and Technology.

Duration: 09/2012 - 08/2015

BORIS - Bar Ordering and Information System

As part of the project, a voice and touch-controlled dialogue and recorder system was developed that recommends and prepares cocktails depending on the context.

The system has been revised and further developed many times. For example, it was given a new pump unit in 2015.

The project was a co-operation project with HEITEC AG Erlangen.

In this context, Stefan Bienk's dissertation focussed on the question of how falls by senior citizens can be predicted using sensor data.

Duration:
Phase 1: 04/2011 - 12/2012
Phase 2: 01/2013 - 06/2014

Funding: BMBF Medical Valley Cluster of Excellence.

The work on the eye-tracking project "Bright Eyes", led by Professor Dr Christian Wolff (Media Informatics Group) and Dr Ludwig Hitzenberger (Information Science Group) and in cooperation with the University of Erlangen, resulted in a prototype system that was able to dynamically adapt its content based on gaze behaviour analyses.

The project, which was initiated by various student groups, focussed on creating an archive of audiovisual study content and information. As part of Knowledgebay, students ultimately created and published over 300 media contributions.

TransRouter - Translation Router (Telematics Applications of Common Interest - Language Engineering)

The aim of the TransRouter project was to design and develop a decision support system for planning translation projects.

In the course of this project, profiles for the description of texts, language data, translation tools and translators were developed as a basis for the decision-making process. Based on this, a decision algorithm was designed and implemented. A graphical interactive user interface supports the selection process. Three prototypes of the system were developed and evaluated through user tests.

TransRouter was funded as project LE4-8345 in the 'Telematics Applications of Common Interest Programme' in the 4th Framework Programme by the Commission of the European Community and the Swiss Ministry of Education and Science from 1998 to 2000.

Members of the consortium: Berlitz, Dublin; CST, Copenhagen; University of Edinburgh; TIM/ETI; University of Geneva; Lernout and Hauspie, Munich; LRC, Dublin; University of Regensburg.
Funding: Commission of the European Communities: DG XIII, Telecommunications. Information Market and Exploitation of Research, Berlitz Ireland Ltd (Berlitz GlobalNet.).

SMMI (Phase I/II/MOST/III/IV) - Mixed Mode Interaction: Subprojects "Voice control concepts for navigation devices and telephony" and "Rhapsody modelling"

• Phase I: Mixed Mode Interaction in complex multi-instrumental and multimodal environments
• Phase II: Development of a prototype for voice control of navigation devices
• MOST: Technical integration of the optical bus system Most for data exchange of various communication components in vehicles
• Phase III: Creation of a natural language dialogue model for the use of hands-free devices in vehicles
• Phase IV: Based on the CASE tool "Rhapsody", which is based on the object-oriented programming paradigm of UML (Unified Modelling Language), a simulation and test environment is created that is able to simulate the entire linguistic and haptic operation and the resulting dependencies of the MMI via the exchange of class libraries. With this tool, code can be generated simultaneously during the creation and modification of the model, which can be used for realisation on other technical platforms.

The focus of this project was therefore the creation of the most general possible class libraries for dialogue description and the integration of general dialogue functionality. The suitability and performance of this design methodology was also to be tested.

Funding: Audi AG, Ingolstadt.

MIROSLAV - Machine Translation Initiative for Russian and other Slavic Languages

The aim of the project was to bring the languages of Eastern Europe closer to the standards and procedures of natural language processing in the rest of Europe. The project comprised two main aspects:

• basic linguistics research: in particular analysing notoriously difficult phenomena of the Slavic languages
• Application relevance and usability of the results obtained: in particular embedding in concrete applications etc.

SIMBA - Linguistic interaction with an intelligent, machine passenger

Design of an interactive acoustic instruction manual in the automotive sector. Prototypical realisation of an interactive acoustic instruction manual in the automotive sector. Experimental investigations into an interactive acoustic operating manual in the automotive sector.

Funding: Daimler-Benz Research, Stuttgart.

ACCESS - Automated Call Centre Through Speech Understanding

Creation of an evaluation tool for the evaluation of dialogues in call centres, analysis of dialogues and Wizzard of Oz experiments for the design of natural language dialogues.

Insurance companies, banks and other service providers are increasingly using call centres to support their customers. For routine dialogues, the ACCESS speech-server project is developing a computer-controlled dialogue with the user that does not differ significantly from operator-controlled dialogues.

Project partners: DASA, Ulm; Daimler Benz, Research Centre, Ulm; Interamerican, Athens; Toshiba, Regensburg; Tellit Direkt Versicherung, Bergisch-Gladbach; University of Patras; WCL, Patras.

Funding: Daimler Benz, Research Centre, Ulm; European Union, Telematics Applications Program No. LE 1802-10347/0; Daimler-Benz Aerospace AG (DASA), Ulm.

WING-IIR - Materials information system with natural language/graphical user interface and intelligent information retrieval

The project focussed on the application field of material fact databases and aimed at the development of an information system with a multimodal user interface and intelligent IR components.

Application partners (selection): Motoren- und Turbinen-Union München GmbH (MTU), Institute for Advanced Materials at the Joint Research Centre of the EC in Petten, Netherlands

Funding: Federal Ministry of Economics (BMWi)

ComfoLIR - ComfoTex & Linguistics Information Science Regensburg

Since 1987, Linguistics Information Science Regensburg (LIR) has devoted itself to several user studies on the evaluation of office software.

As part of the ComfoLIR research project, the user interface of Siemens' word processing programme ComfoTex was evaluated and ComfoHelp, an intelligent help system for the software, was developed.

Funding: Siemens Nixdorf Information Systems (SNI)

DICOS - Dialogues with computers in natural language

The project dealt with human language behaviour in the context of human-machine interaction. Empirical evidence for differences in language use between human-machine dialogue and interpersonal communication was collected as part of several simulation studies in the area of library and railway information.

PADOK - Patent documentation

For the development of a German patent information system, several content indexing procedures were subjected to retrieval and evaluation tests as part of the project.

The computational linguistics analysis of the patent specifications was partly carried out in cooperation with the German Patent Office.

The project was divided into two phases:

PADOK-I: 1984 - 1986

PADOK-II: 1986 - 1989

Funding: Federal Ministry of Research and Technology (BMFT)

COBIS - Computerised Office Information System

As part of the project, a laboratory test was carried out in which CONDOR, a natural language information retrieval system from Siemens, was used for the first time in realistic office situations. Among other things, the project focused on increasing the efficiency of office work and on the use of dialogue-based information systems to improve communication processes at the workplace.

Funding: Federal Ministry of Research and Technology (BMFT), Daimler-Benz AG in Stuttgart-Fellbach

JUDO - Legal document analysis

The project focussed on the exemplary development of software for the linguistic processing of legal texts.

Funding: Federal Ministry of Research and Technology (BMFT)

COBAPH - COBOL basic programme for philology

The project focussed on the development of the word processing system COBAPH, which was used to create alphabetical and frequency-based word directories.