Publikationen

Autor	Frick, Raphael Antonius; Vogel, Inna
Datum	2022
Art	Conference Paper
Abstrakt	During the corona pandemic misinformation has been increasingly spread on social media. Since the automatic verification of social media postings has shown to be challenging, there exists the need of systems, that can identify whether a claim in a post has already been previously analyzed by independent fact-checkers. In this paper, a system based on ensemble classification is proposed. It takes advantage of state-of-the-art sentence transformers for estimating the semantic similarity between a given tweet and individual parts of a fact-check. Furthermore, it incorporates several preprocessing steps as well as back-translation as a data augmentation technique. The proposed model ranked sixth best in the competition.
Konferenz	Conference and Labs of the Evaluation Forum 2022
Url	https://publica.fraunhofer.de/handle/publica/427143

Fraunhofer SIT at CheckThat! 2022: Ensemble Similarity Estimation for Finding Previously Fact-Checked Claims