We combine teaching with research. This gives our students the opportunity to work on the latest topics around technology, information, and people. We also actively support our students in publishing their work during their studies.
In the context of ECIR 2023, the Workshop on Reducing Online Misinformation through Credible Information Retrieval (ROMCIR) took place. Hoai Nam Tran presented his paper "Towards Reducing Misinformation and Toxic Content Using Cross-Lingual Text Summarization", which is available as a presentation and will be published in the proceedings of the workshop.
At CHIIR 2023, which took place this year on-site in Austin (Texas, USA), Markus Bink presented his paper "Investigating the Influence of Featured Snippets on User Attitudes". The paper was written together with Sebastian Schwarz (Master student Media Informatics) and was supervised by David Elsweiler and Tim Draws (Delft University of Technology).
Maximilian Weissenbacher's paper on "Sentiment Analysis on Twitter for the Major German Parties during the 2021 German Federal Election" was presented at KONVENS 2022. This research looked at how the sentiment of political parties changes on Twitter during an election period. This involved using text classification methods and comparing their performance, finding that transformer-based models such as the bidirectional encoder of transformers (BERT) performed better than traditional machine learning models such as Naive Bayes and lexicon-based models such as GerVADER. The paper was written in collaboration with Thomas Schmidt (who presented at KONVENS 2022), Jakob Fehle, Jonathan Richter, Philipp Gottschalk, and Christian Wolff.
Hoai Nam Tran was very successful in Task 3B of the CheckThat! Lab at CLEF 2022 and won first place with his approach. The topic was "Fighting the COVID-19 Infodemic and Fake News Detection". The focus was on Cross-lingual Fake News Detection and was extremely challenging due to the composition of English training data and German test data, whose extracted news articles comprised up to 100,000 characters. Out of 16 participating teams, only 8 managed a correct submission and even here Hoai Nam Tran was able to stand out clearly (6.5%) from the runners-up with a Macro-F1 score of 0.290 and an Accuracy of 0.427.
His approach consisted of three parts: Summarisation Technique, Machine Translation, and Multi-Class Classification. The results have also been published as a CLEF 2022 paper.
Miriam Schirmer, together with Gregor Donabauer and Udo Kruschwitz, published the paper "A New Dataset for Topic-Based Paragraph Classification in Genocide-Related Court Transcripts", which was accepted for LREC 2022 and presented as a poster. Miriam Schirmer studied Criminology in her Master's program at UR and attended a course on Natural Language Engineering, which had a lasting impact on her. At the same time, the paper is part of her PhD in Computational Social Science at TU Munich.
Philipp Hartl worked together with Udo Kruschwitz on the paper "Applying Automatic Text Summarization for Fake News Detection", which was accepted for LREC 2022. The paper is based on Philipp Hartl's master thesis.
Christoph Turban, together with Udo Kruschwitz, submitted the paper "Tackling Irony Detection using Ensemble Classifiers" for LREC 2022. The paper was accepted. The basis for this paper was the bachelor thesis of Christoph Turban.
Markus Bink submitted the paper "Featured Snippets and their Influence on Users' Credibility Judgements" for CHIIR 2022, which has already been accepted for the conference. The publication will follow shortly. The paper was written in the context of his bachelor thesis and was realized together with Steven Zimmerman and David Elsweiler.
Gregor Donabauer, in collaboration with Prof. Dr. Bernd Ludwig, published his master thesis entitled "Testing Landmark Salience Prediction in Indoor Environments Based on Visual Information" as a short paper at GIScience 2021.
Our postgraduate students Hoai Nam Tran and Maximilian Schmidhuber each successfully participated as one-person teams in the GermEval 2021 Shared Task. The subtasks posed challenges to the detection or classification of toxic hate comments on the one hand, and of valuable comments and facts on the other hand.
Hoai Nam Tran took 2nd place overall out of 15 teams. Maximilian Schmidhuber completed the first subtask and placed 9th there. Their performance is particularly noteworthy because they competed as one-person student teams against group teams of PhD students, postdocs and industry representatives.
Our students' approaches for the Shared Task were published in the respective proceedings .
In collaboration with Prof. Dr. Kruschwitz, Gregor Donabauer prepared the paper "University of regensburg@ swisstext 2021 sepp-nlg: Adding sentence structure to unpunctuated text". It was written as part of a shared task at this conference and is related to his presentation/project at WWW'21.
At WWW 2021, Gregor Donabauer (together with Udo Kruschwitz and David Coney) presented a paper on "Making Sense of Subtitles: Sentence Boundary Detection and Speaker Change Detection in Unpunctuated Texts". This project was part of the seminar "Advanced Topics in Information Retrieval".
At ECIR 2021, which takes place in an online format, Dominik Ramsauer presented a short paper on "Exploring the Incorporation of Opinion Polarity for Abstractive Multi-document Summarization". In it, he elaborated a model for MDS that makes it possible to also obtain sentiments and opinions within abstract summarized documents.
Anna-Marie Ortloff presented the paper "The Effect of Nudges and Boosts on Browsing Privacy in a Naturalistic Environment" at CHIIR 2021, which deals with ways to influence user behavior regarding the disclosure of personal data while browsing the internet. This work emerged from her master thesis and was carried out in collaboration with Steven Zimmerman, David Elsweiler and Niels Henze.
Thomas Jänich, graduate of Information Science in Regensburg, started his PhD at the University of Glasgow in February 2021. He is working on the topic of "Fair Machine Learning for Search & Recommendation Systems".
Anna-Marie Ortloff, former student of Information Science at the University of Regensburg, started her PhD at the University of Bonn in early January 2021. As part of the Workgroup Behavioral Security (BeSec) under Prof. Dr. Matthew Smith, she would like to find her PhD topic in this area, considering Human Information Behavior.
Julia Hoffmann published (together with Udo Kruschwitz) the paper "UR NLP @ HaSpeeDe 2 at EVALITA 2020: Towards Robust Hate Speech Detection with Contextual Embeddings". The paper originated on the edited Shared Task of the conference, which in turn was initiated in the Master Project Course. On the topic of Hate Speech Detection, she also participated in the workshop, as well as the Women in Data Science 2021 conference.
Sabrina Barko-Sherif's paper "Conversational Agents for Recipe Recommendation" was accepted for and presented at CHIIR '20 (The 2020 ACM Conference on Human Information Interaction and Retrieval).
Selina Meyer published the results of her master's thesis as an article "Assessing the Quality of Weight Loss Information on the German Language Web" in the journal Movement and Nutrition in Health and Disease.
At the end of 2015 six student groups from our master course had the chance to present their projects at the MUM’15 ('International Conference on Mobile and Ubiquitous Multimedia') in Linz. We are proud to say that 6 of the 15 accepted projects came from our students.
Below is a listing of those projects:
Nombot is a dialog system which makes it easier to track your dietary habits (to the paper).
FROY is a mobile movie recommendation system which matches users emotional state with the mood of movies (to the paper).
This project tries to answer the question on how a smartwatch with speech recognition can be used to win landmarks for navigational systems (to the paper).
This project aimed to optimize destination input on the campus intern pedestrian navigation system "URwalking" (to the paper).
The goal was to implement an application which allows the user to choose songs according to his or her mood (to the paper).
An augmented reality-based assistance system was developed in this project. It supports industry workers in the fabrication process of circuit panels with a smart glass and a self-developed software (to the paper).
Students conduct an eye tracking experiment.