Few-Shot (FS) Learning for Automatic content analysis in Communication studies (FLACA)
The FLACA consortium project aims to make the latest advances in the field of Natural Language Processing in computer science in relation to Few-Shot (FS) Learning accessible for automatic content analysis in communication science. In recent years, two areas of computer science, which have enormous potential for the (partial) automation of content analysis (ACA) – and thus also for increasing the data competence of communication science as a whole – have seen extensive growth. On the one hand, pre-trained language models based on neural transformer networks and subsequent Few-Shot text classification enable reliable identification of content categories with relatively little training data. On the other hand, argument mining methods allow the automatic coding of argument components and positions. These developments address the central desideratum of contemporary content analysis research: The evaluation of very large amounts of text for semantically complex categories.
The necessary procedure and technical implementation of the project’s goals will be developed in the context of several case studies, e.g., the framing of arms deliveries to Ukraine in German media reporting. The project will provide scientific publications, best practices, software, and e-learning resources to help communication scientists access and further develop these new technologies from computer science according to the requirements of their discipline. For the transfer of the new methods, FLACA places a main focus on young academics, for whom data competencies for ACA will be taught within the framework of methods workshops.
FLACA is a consortium project with the Leibniz Institute for Media Research Hans-Bredow-Institute.
The project is funded from the research plan Data Action of the Federal Ministry of Education and Research. The funding measure for "Strengthening the Data Competences of Young Scientists" aims to expand and deepen the data competence of young scientists at universities and non-university research institutions in the diverse fields of the scientific landscape through the linking of specialized data science skills with field-specific knowledge. The projects are financed from the building and resilience facility of the EU recovery plan "NextGenerationEU."
The FLACA-team:
UHH: Prof. Dr. Katharina Kleinen-von Königslöw, Dr. Gerret von Nordheim, Dr. Kostiantyn Yanchenko
Consotium partners /HBI: Dr. Gregor Wiedemann, Dr. Jonas Rieger, Mattes Ruckdeschel
- Duration: 2022-2025
- Project lead: Prof. Dr. Katharina Kleinen-von Königslöw
- Sponsor: BMBF