Apprenticeship

• Merging ecocritical discourses with approaches from sustainability studies
• Individual case studies of a selection of
• extinct species (e.g. the bear)
• returning or reintroduced species (wolf, lynx, beaver)
• invasive species

Method:

• Use of an ML tool for the automatic annotation of animals, plants and habitats
• Application to a large, diachronic corpus
• Analyses of typical narrative patterns and quantitative processes
• Practical component: Development of an explorative (prototypical) presentation on a website

Contact person: Mareike Schumacher

AI-based voice models and voice cloning processes are increasingly being used in the dubbing of films, series and digital media. In some cases, dubbing actors are being asked to provide their voices for the training of corresponding AI systems (see the statement by the Association of German Voice Actors (external link, opens in a new window) or the call by individual voice actors (external link, opens in a new window) ). This development has implications for labour and copyright law as well as fundamental cultural studies consequences with regard to voice, authenticity and perception. From the perspective of digital humanities, it is of particular interest how recipients react to AI-generated voices and whether human and artificial synchronisation can be distinguished in reception.

This MA project addresses these questions as part of an empirical reception study using digital methods. The aim is to systematically analyse whether recipients can distinguish AI-generated dubbing voices from human speakers and how they are evaluated in each case. The perception of naturalness, credibility and emotional appropriateness of the voices as well as the preferences expressed by the recipients are analysed. A central aspect of the study is the comparison of reception situations in which it is known that the voices are AI-generated with those in which this information is not available.

As part of the study, a digital audio corpus with short synchronised sequences will be created, which includes both human-recorded and AI-generated versions of comparable content. The evaluation will be based on standardised surveys and statistical analysis methods. The aim is to make well-founded statements about the perception, acceptance and aesthetic evaluation of AI-supported synchronisation on a reliable data basis and to classify these in current cultural studies and media theory discourses in the digital humanities.

Contact person: Clara Helmig

In this MA project, a prototype is to be developed with which texts can be annotated quickly, easily and using mobile technologies (mobile first) and gamification approaches. Users must first pass an annotation training course in which a text is annotated using annotation guidelines stored in the tool. If the user's annotation corresponds to at least 80% of a gold standard annotation (IAA 0.8), the actual annotation area is activated. Here, users can select and annotate texts stored in the backend according to their preferences. They can win various awards and view statistics on their annotations.

While the annotation front end is primarily intended to be used on mobile devices, the back end behind it can be operated on a PC. Here it must be possible for project managers and employees to store texts in annotation format (txt, xml, csv, tsv, possibly also docx and pdf). The descriptions of the texts that are displayed to users and keywords for the respective texts can also be entered here. Descriptions and keywords can then also be used as the basis for the annotators' statistics. The application should be optimised for the following usage scenarios:

1. Use in research teams
   In this usage scenario, the focus is on being able to summarise and export the various complete annotations of a text in the form of gold standard annotations. In this scenario, project managers operate the back end and users operate the front end.
2. Use in university teaching
   In this usage scenario, the focus is on ensuring that the front-end application can be used with a low threshold and without a lengthy introduction. It is also important that users annotate in a protected login area. In this scenario, users operate both the backend and the frontend and thus learn to deal with multiple roles.

Expected prior knowledge: Good programming skills (especially web applications), basic knowledge of annotation techniques and workflows, interest in gamification

Contact person: Mareike Schumacher

Further sources:

• Artstein, R. (2017) 'Inter-annotator Agreement', in N. Ide and J. Pustejovsky (eds) Handbook of Linguistic Annotation. Dordrecht: Springer Netherlands, pp. 297-313. available at: https://doi.org/10.1007/978-94-024-0881-2_11.
• Artstein, R. and Poesio, M. (2008) 'Inter-Coder Agreement for Computational Linguistics', Computational Linguistics, 34(4), pp. 555-596. Available at: https://doi.org/10.1162/coli.07-034-R2.
• Bögel, T. et al. (2015) 'Collaborative Text Annotation Meets Machine Learning: heureCLÉA, a Digital Heuristic of Narrative'. Available at: https://doi.org/10.5281/ZENODO.3240591.
• Frey-Endres, M. and Simon, T. (2021) Digital tools for text-based annotation, corpus analysis and network analysis in the humanities. Edited by S. Bartsch et al. Darmstadt. Available at: tuprints.ulb.tu-darmstadt.de/17850/ (Accessed: 7 April 2021).
• Pagel, J. et al. (2020) 'Annotation als flexibel einsetzbare Methode', in N. Reiter, A. Pichler, and J. Kuhn (eds) Reflektierte algorithmische Textanalyse. De Gruyter, pp. 125-142. available at: https://doi.org/10.1515/9783110693973-006.
• Reiter, N. (2020) 'Anleitung zur Erstellung von Annotationsrichtlinien', in Reiter, Nils, A. Pichler, and J. Kuhn (eds) Reflektierte algorithmische Textanalyse. Berlin: De Gruyter, pp. 193-202. available at: https://doi.org/10.1515/9783110693973-009.
• Reiter, N. and Konle, L. (2022) 'Measurement methods for inter-annotator-agreement (IAA)'. Available at: https://doi.org/10.47952/gro-publ-103.
• Wissler, L. et al. (2014) The Gold Standard in Corpus Annotation. Available at: doi.org/10.13140/2.1.4316.3523.

In this MA project, a prototype is to be developed with which images can be annotated quickly, easily and using mobile technologies (mobile first) and gamification approaches. Users must first pass an annotation training course in which a number of images are annotated using guidelines stored in the tool. If the user's annotation corresponds to at least 80% of a gold standard annotation (IAA 0.8), the actual annotation area is activated. Here, users can select and annotate images stored in the backend according to their preferences. They can win various awards and view statistics on their annotations.

While the annotation front end is primarily intended to be used on mobile devices, the back end behind it can be operated on a PC. Here it must be possible for project managers and employees to store images in annotation format (png, jpeg, pdf). The descriptions of the images that are displayed to users and keywords for the respective images can also be entered here. Descriptions and keywords can then be used as the basis for the annotators' statistics. The application should be optimised for the following usage scenarios:

1. Use in research teams
   In this usage scenario, the focus is on being able to summarise and export the various complete annotations of an image in the form of gold standard annotations. In this scenario, project managers operate the back end and users operate the front end.
2. Use in university teaching
   In this usage scenario, the focus is on ensuring that the front-end application can be used with a low threshold and without a lengthy introduction. It is also important that users annotate in a protected login area. In this scenario, users operate both the backend and the frontend and thus learn to deal with multiple roles.

Expected prior knowledge: Good programming skills (especially web applications), basic knowledge of annotation techniques and workflows, interest in gamification

Contact person: Mareike Schumacher

Further sources:

• Artstein, R. (2017) 'Inter-annotator Agreement', in N. Ide and J. Pustejovsky (eds) Handbook of Linguistic Annotation. Dordrecht: Springer Netherlands, pp. 297-313. available at: https://doi.org/10.1007/978-94-024-0881-2_11.
• Artstein, R. and Poesio, M. (2008) 'Inter-Coder Agreement for Computational Linguistics', Computational Linguistics, 34(4), pp. 555-596. Available at: https://doi.org/10.1162/coli.07-034-R2.
• Bögel, T. et al. (2015) 'Collaborative Text Annotation Meets Machine Learning: heureCLÉA, a Digital Heuristic of Narrative'. Available at: https://doi.org/10.5281/ZENODO.3240591.
• Frey-Endres, M. and Simon, T. (2021) Digital tools for text-based annotation, corpus analysis and network analysis in the humanities. Edited by S. Bartsch et al. Darmstadt. Available at: tuprints.ulb.tu-darmstadt.de/17850/ (Accessed: 7 April 2021).
• Pagel, J. et al. (2020) 'Annotation als flexibel einsetzbare Methode', in N. Reiter, A. Pichler, and J. Kuhn (eds) Reflektierte algorithmische Textanalyse. De Gruyter, pp. 125-142. available at: https://doi.org/10.1515/9783110693973-006.
• Reiter, N. (2020) 'Anleitung zur Erstellung von Annotationsrichtlinien', in Reiter, Nils, A. Pichler, and J. Kuhn (eds) Reflektierte algorithmische Textanalyse. Berlin: De Gruyter, pp. 193-202. available at: https://doi.org/10.1515/9783110693973-009.
• Reiter, N. and Konle, L. (2022) 'Measurement methods for inter-annotator-agreement (IAA)'. Available at: https://doi.org/10.47952/gro-publ-103.
• Wissler, L. et al. (2014) The Gold Standard in Corpus Annotation. Available at: https://doi.org/10.13140/2.1.4316.3523.

As part of the research project "From Epistemic Injustice to Epistemic Awareness. Process-based methodological research on abuse of adult women in the Catholic Church"(EpiEpa (external link, opens in a new window)), students have the opportunity to write their final theses in the Master's programme in Digital Humanities.

The topics range from annotation, qualitative and hermeneutic analysis to technical DH application areas such as computer-aided text analysis and AI-supported methods.

In the project, we work with different types of texts:

• autobiographical texts by those affected
• Media reports / web texts
• Studies by the dioceses
• Files of the dioceses
• Social media data

Objectives from a DH perspective include

1. Relevance detection: identification of documents and text passages on the topic of abuse of adult women, under conditions of massive class imbalance and epistemic ignorance.
2. Classification tasks: Prediction of annotated classes on broader topics (e.g. framing, actions, narratives).
3. Application of methods: Use of computer-aided methods (e.g. LLMs, sentiment analysis, topic modelling) to gain in-depth knowledge on relevant and non-relevant data.
4. Annotation and annotation analyses of different text types.

Theses are possible in all these areas and for all text types. Exemplary topics:

• Evaluation of LLMs for relevance detection in abuse studies
• Fine-grained classification of a complex annotation scheme in autobiographical texts
• Comparison of different relevance detection approaches (semantic retrieval, LLMs, discriminative transformers) in web texts
• Computer-aided text analysis of media reports on gottes-suche.de with a focus on abuse of adult women vs. other affected groups
• Social media analyses, e.g:
  • Discourse analysis of Christian subreddits
  • Analysis of #NunsToo in the social media context

Your own suggestions for topics in the project context are expressly welcome.

If you are interested, please contact me:
Thomas Schmidt: thomas.schmidt​(at)​ur.de (opens your email program)

Based on the idea of a tripartite division of gender into gender, gender identity and gender performance, this work focusses on the aspect of doing gender in literary texts. In contrast to gender and gender identity, gender performance is a consciously chosen strategy by means of which gender stereotypes can be taken up, rejected or broken. In literary texts, both typical behaviours of characters of different genders and unusual ways of negotiating that break with a binary understanding of gender can be depicted. In a quantitative approach, this thesis uses a corpus of literary texts to identify typical, i.e. frequent, behaviours of male and female characters. To this end, text mining methods are combined in such a way that characters and verbs are first identified and then brought together using coreference resolution. The corpus can be chosen individually, but should be designed as a balanced collection of texts in which both binary and non-binary gender figures occur.

Expected prior knowledge: In-depth knowledge of text mining and basic programming skills should be available.

Contact person: Mareike Schumacher

Further sources:

• Butler, Judith 1956-. The discomfort of gender. 1. Edition, [reprint]. Edition Suhrkamp: Gender studies: On the difference between the sexes. Frankfurt am Main: Suhrkamp, 2003.
• Jockers, Matthew, and Gabi Kirilloff. "Understanding Gender and Character Agency in the 19th Century Novel". Journal of Cultural Analytics 2, no. 2 (1 December 2016). doi.org/10.22148/16.010.
• Baylog, Oliver, Laura Dimmit, T. Heller, Gabi Kirilloff, Shannon Smith, G. Thomas, Chandler Warren, and James Wehrwein. "" More than Custom has Pronounced Necessary " : Exploring the Correlation between Gendered Verbs and Character in the 19th Century Novel Nebraska Literary Lab," 2016. www.semanticscholar.org/paper/%22-More-than-Custom-has-Pronounced-Necessary-%E2%80%9D-%3A-the-Baylog-Dimmit/631b70f5581d6df8b1e35f9f679465a119b499ba.

Description: In the context of personal scientific communication or in private, researchers use social media to present themselves and/or their research. Following the narrative identity thesis (Ricoeur) and approaches of sociolinguistic (Georgakopoulou, Bamberg) and multimedia narratology (Page, Ryan), this paper explores the question of how researchers present themselves and which narrative identities they form. Special attention is paid to the (self-)representation of female scientists: Do women portray themselves and their everyday working lives differently from people who identify with other gender forms? The questions will be analysed using either a predefined Twitter or Instagram corpus. A relevant subset will be extracted from the entire corpus and analysed. Methods of text mining and digital social media analysis will be applied.

Expected prior knowledge: Knowledge of digital social media analysis and basic knowledge of coding should be available.

Contact person: Mareike Schumacher

Further sources:

• Bamberg, Michael. "Identity and Narration". In the living handbook of narratology, 2019.
• Bamberg, Michael. "Narrative Discourse and Identities". In Narrative Discourse and Identities, 213-38. De Gruyter, 2008. doi.org/10.1515/9783110201840.213.
• Georgakopoulou, Alexandra. Small Stories, Interaction and Identities. Sin.8. John Benjamins Publishing Company. Accessed 1 June 2023. benjamins.com/catalog/sin.8.
• Page, Ruth. Narratives Online: Shared Stories in Social Media. Cambridge University Press, 2018.
• Ricoeur, Paul. "Narrative identity". In Heidelberger Jahrbücher, edited by Elmar Mittler, 57-67. Heidelberger Jahrbücher. Berlin, Heidelberg: Springer, 1987. doi.org/10.1007/978-3-642-71777-2_5.
• Ryan, Marie-Laure, James Ruppert, and John W. Bernet. Narrative Across Media: The Languages of Storytelling. U of Nebraska Press, 2004.

Can tools like ChatGPT really imitate the style of famous authors? Will our literature only be artificially generated in the future? And do readers even want to read history that wasn't written by humans? The rapid development of AI tools such as ChatGPT can trigger concerns about the literary scene, especially if the assumption that text-generating AIs (genAI or genKI) can perfectly imitate human writing (cf. Bajohr and Roloff 2024) is adopted without scrutiny.
In the project The Digital Poet, artificially generated fairy tales in which mermaids play a role are published with different GPT versions in order to show what such literature looks like today and to what extent it improves from version to version. This MA project will now systematically investigate the following as part of a reception study

1. Which texts are recognised by readers as AI-generated texts and how well, and
2. How similar the author's style is rated in comparison with the model author
   a. if readers know that the texts are AI-generated texts and
   b. if readers do not know that the texts are AI-generated texts.

The primary text corpus of the digital poet and a comparison corpus with mermaid narratives written by humans will be made available for the study. The aim of the study is to answer the research questions on the basis of reliable data. Publication of the work is desirable in the event of an excellent evaluation.

Expected basic knowledge: Great interest in and good performance in survey-based project work, interest in literary studies (reception aesthetics).

Contact person: Mareike Schumacher

Further sources:

• Bajohr, H. (2022) Dumb Meaning: Machine Learning and Artificial Semantics. Available at: https://doi.org/10.13140/RG.2.2.13207.65444/1.
• Bajohr, H. (2025) Thinking <em>with</em> AI: Machine Learning the Humanities. Open Humanites Press. Available at: www.openhumanitiespress.org/books/titles/thinking-with-ai/ (Accessed: 22 April 2025).
• Bajohr, H. and Roloff, S. (2024) Digitale Literatur zur Einführung. Hamburg: Junius Verlag (Introduction).
• Chakrabarty, T. et al. (2024) 'Art or Artifice? Large Language Models and the False Promise of Creativity', in Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems. New York, NY, USA: Association for Computing Machinery (CHI '24), pp. 1-34. Available at: https://doi.org/10.1145/3613904.3642731.
• Mirowski, P. et al. (2024) 'A Robot Walks into a Bar: Can Language Models Serve as Creativity SupportTools for Comedy? An Evaluation of LLMs' Humour Alignment with Comedians', in Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency. New York, NY, USA: Association for Computing Machinery (FAccT '24), pp. 1622-1636. available at: https://doi.org/10.1145/3630106.3658993.
• Rettberg, S. (2023) 'Cyborg Authorship: Writing with AI - Part 1: The Trouble(s) with ChatGPT'. Available at: https://doi.org/10.7273/5SY5-RX37.
• Rettberg, S. and Rettberg, J.W. (2024) 'Algorithmic narrativity: Literary experiments that drive technology', Dialogues on Digital Society, p. 29768640241255848. Available at: https://doi.org/10.1177/29768640241255848.
• Vlaad, S. (2025) 'Texts without authors: ascribing literary meaning in the case of AI', The Journal of Aesthetics and Art Criticism, 83(1), pp. 4-11. Available at: doi.org/10.1093/jaac/kpae047.

Using a corpus of digitised travel reports by the Swiss poet and publicist Franz Xaver Bronner, two prototype applications are to be developed and evaluated comparatively from a usability perspective. The aim is to make it possible to experience the travel routes interactively. To this end, relevant categories of the spatial representation should first be automatically or semi-automatically annotated in a text mining process and then transferred to a geovisualisation. In comparison with digital editions of other travel reports such as Humboldt digital or the literary atlas of Europe (Piatti), the state of the art of cultural-geographical visualisations should be taken into account and, if necessary, supplemented by innovative aspects. Graphical representations that deviate from classic geovisualisation are also conceivable, allowing a cultural-sociological view of the texts rather than a geographical one, for example.

Expected prior knowledge: basic knowledge of text mining and in-depth knowledge of geovisualisation desirable

Contact person: Mareike Schumacher (the work will also be supervised by Anna Ananieva from the DeHisRe project of the Institute for East and South East European Studies)

further sources:

• Piatti, Barbara 1973-. The geography of literature: settings, spaces of action, spatial fantasies. Wallstein, 2008.

The writer and journalist Franz Xaver Bronner published a series of travel reports in the late 18th and early 19th century that still provide a glimpse of the Europe of his contemporaries today. He was also in lively dialogue with the greats of his time, such as Christoph Martin Wieland. Based on the question of which spaces are described by Bronner and with which cultural aspects these are linked, different text types and document types will be analysed and visualised. (Historical) maps can be combined with pure text in the form of reports and letters as well as digital facsimiles. The scientific examination of spatial intertextual references, such as recurring spatial narratives in travel reports and letters, should form the basis of a knowledge graph, which in turn could be part of a web presentation of a Bronner edition. Special consideration should be given to the inclusion of Linked Open Data and common standard data as well as the common standard database (GND).

Expected prior knowledge: Experience with text annotation, linked open data and graph databases desirable

Contact person: Mareike Schumacher (the work will also be supervised by Sandra Balck from the DeHisRe project of the Institute for East and South East European Studies)