Automatic Swabian Recognition | University of Tübingen

Automatic Swabian Recognition (ASR): Processing and Evaluating Dialect Speech

Gefördert von	Innovation Fund Program of the Cluster of Excellence „Machine Learning: New Perspectives for Science”
Projektleitung	Prof. Dr. Thomas Thiemeyer, Prof. Dr. Lea Frermann (Seminar für Sprachwissenschaft), Prof. Dr. Gerhard Jäger (Seminar für Sprachwissenschaft), Prof. Dr. Kerstin Ritter (Hertie Institute for AI in Brain Health)
Projektmitarbeiter*innen	Dr.ⁱⁿ Valeska Flor, NN, NN
Projektlaufzeit:	2027 - 2029 (+1 Jahr)

The goal of Automatic Swabian Recognition (ASR) is to make the Arno Ruoff archive fully usable for both the scientific community and the wider public by creating a curated, open dataset with parallel tiers of (i) sound recordings, (ii) existing transcriptions in German orthography, and (iii) phonetic transcriptions in IPA produced by a dialect-adapted automatic speech recognition pipeline.

Developing robust automatic speech recognition models on archival dialect data enables transfer to contemporary applications facing similar challenges: regional variation, spontaneous speech, and limited training data. The project addresses a critical gap: current automatic speech recognition systems fail on regional dialects, limiting cognitive assessment for neurodegenerative diseases. By validating dialect-aware models on the TREND cohort (1,200 participants, Tübingen region, longitudinal recordings), we establish speech as a non-invasive digital biomarker while ensuring robustness across patient populations.

Building on this dual foundation of historical dialect archives and contemporary clinical speech data, the project integrates machine-learning methods, quantitative models of linguistic change, and qualitative methods of cultural-anthropological analyses of archival meaning and access.

The project “Automatic Swabian Recognition (ASR): Processing and Evaluating Dialect Speech” is a network project funded by the Innovation Fund of the Cluster of Excellence “Machine Learning: New Perspectives for Science” at the University of Tübingen. Its core resource is the Arno Ruoff Archive, around 1,600 interviews with speakers of Upper German dialects recorded between 1950 and 1975. The project develops a dialect-adapted speech-recognition and IPA-transcription pipeline for this material and releases it as an open benchmark dataset, uniting computational linguistics, machine learning, and cultural anthropology.

Within this framework, we are seeking two fill two doctoral researcher positions. Each doctoral researcher will pursue a dissertation within the project.

Position 1: Dialect-aware automatic speech recognition
Jointly supervised by Prof. Dr. Gerhard Jäger (Department of Linguistics) and Prof. Dr. Kerstin Ritter (Hertie Institute for AI in Brain Health)
State-of-the-art speech recognition degrades sharply on regional dialects, historical recordings, and spontaneous speech — precisely the conditions of the Arno Ruoff Archive. Building on self-supervised speech representations, this dissertation develops uncertainty-aware, low-resource ASR and IPA transcription for dialect speech, with an open benchmark dataset as its central deliverable. The resulting methods are also of interest for clinical purposes, where speech is increasingly explored as a non-invasive marker of cognitive change and neurodegeneration.

Position 2: Modeling Narrative Knowledge in Cultural Context
Jointly supervised by Prof. Dr. Lea Frerman (Department of Linguistics), Dr. Valeska Flor and Prof. Dr. Thomas Thiemeyer (Ludwig-Uhland-Institut für Empirische Kulturwissenschaft)
Situated at the intersection of NLP and historical and cultural anthropology, the project examines how narratives and cultural knowledge in historical oral archives are shaped by their original recording contexts and by contemporary digital processing. This dissertation will leverage the transcripts of the Arno Ruoff Archive to develop novel computational models and representations of cultural narratives which are adapted to historical, underrepresented cultures thereby contributing foundational CL/ML methodology for context-aware narrative understanding with applications in linguistic and cultural research and archive exploration.

Your profile
We are looking for candidates with a background in machine learning, computer science, computational linguistics, data science, cultural anthropology, or a related field. Beyond strong technical skills — in particular solid programming experience in Python and a deep-learning framework such as PyTorch — we especially value a genuine interest in interdisciplinary work, adaptability, creativity, and the willingness to delve into unfamiliar methods, materials, and disciplinary perspectives. Knowledge of German is desirable for in-depth engagement with the archival materials. For Project 1, experience with speech or audio processing, self-supervised learning, or sequence models is a plus.

Conditions and application
The salary follows the union contract TV-L, E13 (75%). The positions are funded for three years, with the possibility of a one-year extension (3+1).
Applications should include a CV, a statement of research interests and relevant experience, and the names of up to two references. The position will be filled as soon as possible; the deadline for applications is September 30th.
Disabled applicants will be preferred if they have the same qualifications as non-disabled applicants. The University of Tübingen strives to increase the proportion of women in research, and therefore encourages qualified female scientists to apply.
Please send your application electronically as a single pdf file to as@semsprach.uni-tuebingen.de, valeska.flor@uni-tuebingen.de and lea.frermann@uni-tuebingen.de.