Element 68Element 45Element 44Element 63Element 64Element 43Element 41Element 46Element 47Element 69Element 76Element 62Element 61Element 81Element 82Element 50Element 52Element 79Element 79Element 7Element 8Element 73Element 74Element 17Element 16Element 75Element 13Element 12Element 14Element 15Element 31Element 32Element 59Element 58Element 71Element 70Element 88Element 88Element 56Element 57Element 54Element 55Element 18Element 20Element 23Element 65Element 21Element 22iconsiconsElement 83iconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsiconsElement 84iconsiconsElement 36Element 35Element 1Element 27Element 28Element 30Element 29Element 24Element 25Element 2Element 1Element 66
Few-Shot Learning for Automated Content Analysis in Communication Science (FLACA)

Few-Shot Learning for Automated Content Analysis in Communication Science (FLACA)

In this project, we apply two methods developed in computer science - Few-Shot Learning and Argument Mining - to the field of empirical communication science. Automated content analysis (ACA) should thus be able to examine even large amounts of data with little coding effort.

The approach required for this as well as the technical implementation will be developed in an exemplary study of positions and argument patterns on Twitter surrounding the Covid 19 pandemic.

As part of the project, we will provide scientific publications, best practices as well as software and e-learning resources that will enable communication science to tap into these new technologies from computer science and develop them further according to their own subject requirements.

For the communication of these new methods, our project focuses on young scientists, who will have the opportunity to acquire data competencies for ACA in methodological workshops.

Cover: Arno Senoner / unsplash
show more

Project Description

Content analysis is one of the central methods of empirical communication science. The ever-increasing amount and availability of digital public communication content makes (partial) automation of content analysis (ACA) imperative.

In computer science, two areas have been intensively researched in the recent past whose current results hold enormous potential for ACA and thus also for increasing data literacy in communication science as a whole: With pre-trained language models based on neural transformation networks (e.g. BERT) and Few-Shot (FS) text classification based on them, it is possible to identify content categories reliably with comparatively little training data. Argument mining methods also enable the automatic coding of argument components and stances.

This addresses two central desiderata of current content analysis research: the evaluation of very large text volumes with semantically complex categories.

Project Information

Overview

Duration: 2022-2025

Research programme:
FP4 - Algorithmed Public Spheres

Third party

BMBF, Projektträger VDI/VDE, Förderkennzeichen 16DKWN064B

Cooperation Partner

Prof. Dr. Katharina Kleinen-von Königslöw, Universität Hamburg

Contact person

Dr. Gregor Wiedemann
Senior Researcher Computational Social Sciences

Dr. Gregor Wiedemann

Leibniz-Institut für Medienforschung | Hans-Bredow-Institut (HBI)
Rothenbaumchaussee 36
20148 Hamburg
Tel. +49 (0)40 450 21 7 55
Fax +49 (0)40 45 02 17 77

Send Email

MAYBE YOU ARE ALSO INTERESTED IN THESE TOPICS?

Newsletter

Subscribe to our newsletter and receive the Institute's latest news via email.

SUBSCRIBE!