Lea Fischbach

Lea Fischbach is a Research Fellow and PhD candidate at the Research Center Deutscher Sprachatlas (DSA) within the Faculty of German Studies and Arts at Philipps University of Marburg. She is employed in the long-term Regionalsprache.de project funded by the Academy of Sciences and Literature in Mainz and also contributes to the BMBF project AnDy. Since January 2021, she has been working at the DSA, initially as a student assistant before advancing to her current PhD candidate position under the supervision of Lucie Flek and Alfred Lameli.

Her educational background includes both Bachelor's and Master's degrees in Informatics from Philipps University of Marburg. Her Bachelor's thesis focused on Efficient Serialization of Objects while her Master's thesis addressed Applied Learning Analytics Process Based on a Language Learning App, building upon an Android app she helped develop as part of a student team project.

Lea's research centers on the classification of German dialects using audio recordings and deep learning techniques. Her work explores how deep learning can be optimized for dialect classification, which phonetic features are crucial for distinguishing between dialects, and how insights from both deep learning and phonetic analysis can be combined for enhanced classification. This research sits at the intersection of computational linguistics, speech processing, and dialectology, with particular focus on German dialects and low-resource audio data scenarios. Her technical approach integrates programming expertise with linguistic knowledge to develop practical solutions for dialect recognition challenges.

Her publications demonstrate a strong focus on practical applications of deep learning to dialect classification challenges. She has explored voice conversion techniques for data augmentation, analyzed the importance of acoustic features in deep learning models, compared speaker diarization approaches for German dialectal speech, and contributed to digital language geography projects. Her work consistently addresses the challenges of working with limited dialectal data and seeks to improve classification accuracy through innovative technical approaches that minimize speaker-related variability while highlighting dialect-specific features.

At Philipps University of Marburg, Lea has served as a tutor teaching courses in programming, technical informatics, and database systems. During her time at the DSA, she has contributed to various projects including the creation of the Hessenplattform and developed the second version of the Welcome to Bavaria app, a language guide for Bavarian dialects. Her technical work bridges the gap between computational methods and traditional linguistic research, creating practical tools for dialect documentation and analysis.

Her research is conducted within the Research Center Deutscher Sprachatlas, a leading institution for German dialect research that provides both the theoretical framework and computational resources for her technical investigations into dialect classification. She actively participates in international conferences including Interspeech, ACL, and NoDaLiDa, presenting her work on deep learning pipelines for dialect recognition and the challenges in developing technical solutions for dialect classification tasks.