Hey, I’m Lea, and my research focuses on the classification of German dialects using audio recordings. I’m currently pursuing my PhD at the Research Center Deutscher Sprachatlas (DSA) at Philipps-Universität Marburg, under the supervision of Lucie Flek and Alfred Lameli. I’m employed in the Regionalsprache.de project, which is a long-term project funded by the Academy of Sciences and Literature in Mainz and also contribute to the BMBF project AnDy. During my time at the DSA, I have worked on system development and contributed to various projects, including the creation of the Hessenplattform.
Since January 2021, I have been working at the DSA. I started as a student assistant and now continue as a PhD candidate. During the time as student assistant, I developed the second version of the app Welcome to Bavaria on my own. The first version was developed as part of a Master’s project in a team of students. Additionally, I have experience as a tutor at Philipps-Universität Marburg, where I taught programming, technical informatics, and database systems.
I completed both my Bachelor’s and Master’s degrees in Informatics at Philipps-Universität Marburg. My Bachelor’s thesis focused on “Effiziente Serialisierung von Objekten” (Efficient Serialization of Objects). For my Master’s thesis, I worked on “Angewandter Learning Analytics Prozess auf Basis einer Sprachlern-App” (Applied Learning Analytics Process Based on a Language Learning App), which was built upon the Android app that I helped develop.
Email: Lea.Fischbach <at> uni-marburg <dot> de
Research Interests
-
Deep Learning for Horizontal Dialect Classification
How can deep learning techniques be optimized to enhance the classification of German dialects from audio recordings? Key aspects include leveraging insights from misclassified segments and optimizing data augmentation strategies for dialect differentiation. -
Phonetic Features in Horizontal Dialect Classification
Which phonetic features are crucial for distinguishing between dialects, and how can they be effectively integrated with deep learning models? This includes evaluating the contribution of vowels and extracting relevant phonetic features. -
Combined Approaches for Enhanced Vertical Dialect Classification
How can insights from both deep learning and phonetic feature analysis be used to develop a more effective horizontal dialect classification model? This involves analyzing speaking behavior to understand code-switching and the impact of social and contextual factors on dialect variation.
Keywords:
Natural Language Processing, Dialectology, Speech Processing, Audio Classification, Acoustic Phonetics, Diatopic and Diaphasic Variation in Dialects, Code-Switching
Publications
-
Lea Fischbach. 2024. A Comparative Analysis of Speaker Diarization Models: Creating a Dataset for German Dialectal Speech. In Proceedings of the 3rd Workshop on NLP Applications to Field Linguistics (Field Matters 2024), pages 43–51, Bangkok, Thailand. Association for Computational Linguistics.
-
Fischbach, Lea, Ganswindt, Brigitte, Lang, Vanessa & Beitel, Dennis. 2024. Dialekte in Hessen. Das Informationsportal zur Sprachgeographie. In: Sprachspuren: Berichte aus dem Deutschen Sprachatlas 4(5). https://doi.org/10.57712/2024-05
Selected talks, poster presentations and other updates
-
“Fallstricke und Herausforderungen bei der Erstellung einer DL-Pipeline zur Klassifikation deutscher Dialekte aus Audiodaten” (Pitfalls and Challenges in Developing a Deep Learning Pipeline for Classifying German Dialects from Audio Data), presentation at the workshop “KI in der Linguistik: Chancen und Herausforderungen (LDDB 2024)” as part of the conference “KI-Methoden im Akademienprogramm: Potenziale und Anwendungsszenarienat”, University of Hamburg (23.9.2024). (Lea Fischbach)
-
“A Comparative Analysis of Speaker Diarization Models: Creating a Dataset for German Dialectal Speech”, presentation at the conference “62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024)”, in the Workshop “Field Matters”, Bangkok, Thailand (16.08.2024). (Lea Fischbach)
-
“EDAudio: Easy Data Augmentation Techniques for Audio Classification”, presentation at the conference “2th International Conference on Language Variation in Europe – ICLaVE|12” in the panel: EMBRACING VARIABILITY IN NATURAL LANGUAGE PROCESSING, Vienna, Austria (10.07.2024). (Alfred Lameli, Lea Fischbach, Caroline Kleen, Akbar Karimi, Lucie Flek)
-
“Dialekte des Deutschen multiperspektivisch betrachtet: Digitale Methoden zur Analyse gesprochener Sprache” (The German dialects from a multiperspective view: Digital methods for the analysis of spoken language), poster at the conference “60. Jahrestagung des Leibniz-Instituts für Deutsche Sprache. Gesprochenes Deutsch: Struktur, Variation, Interaktion”, Mannheim, Germany (06.03.2024). (Lea Fischbach, Marina Frank, Caroline Kleen)
-
„Un programma di apprendimento di una lingua regionale in forma di App: Implementazione didattica, utilizzo e realizzazione tecnica“ (A language learning program for a regional language in the form of an app: Educational implementation, usage, and technical development), presentation at Università degli Studi die Modena e Reggio Emilia (03.10.2022). (Peter Kaspar, Lea Fischbach)
-
„Apps am Sprachatlas – Modernisierungen und Neuentwicklungen“ (Apps at the Research Center Deutscher Sprachatlas – Modernizations and New Developments”), presentation at the workshop “Herausforderungen linguistischer Datenvisualisierung (LDDB 2022)” at University of Vienna (16.9.2022). (Lea Fischbach, Robert Engsterhold)
-
„Welcome to Bavaria – Eine Sprachführer-App für die bayerischen Dialekte“ (Welcome to Bavaria – A language guide app for Bavarian dialects), presentation at the conference „Minderheitensprachen im digitalen Zeitalter. Sprachgebrauch, Spracherhalt, Sprachvermittlung“ at the Alfred Krupp Wissenschaftskolleg Greifswald (12.12.2020). (Lea Fischbach, Hanna Fischer, Milena Gropp, Jeffrey Pheiff)
Contact
Feel free to connect with me through the following platforms or check out any of my profiles: