This paper presents medBERTde, a pre-trained German BERT model specifically designed for the German medical domain. The model has been trained on a large corpus of 4.7 Million German medical documents and has been shown to achieve new state-of-the-art performance on eight different medical benchmarks covering a wide range of disciplines and medical document types. In addition to evaluating the overall performance of the model, this paper also conducts a more in-depth analysis of its capabilities. We investigate the impact of data deduplication on the model's performance, as well as the potential benefits of using more efficient tokenization methods. Our results indicate that domain-specific models such as medBERTde are particularly useful for longer texts, and that deduplication of training data does not necessarily lead to improved performance. Furthermore, we found that efficient tokenization plays only a minor role in improving model performance, and attribute most of the improved performance to the large amount of training data. To encourage further research, the pre-trained model weights and new benchmarks based on radiological data are made publicly available for use by the scientific community.
@article{arxiv.2303.08179,
title = {MEDBERT.de: A Comprehensive German BERT Model for the Medical Domain},
author = {Keno K. Bressem and Jens-Michalis Papaioannou and Paul Grundmann and Florian Borchert and Lisa C. Adams and Leonhard Liu and Felix Busch and Lina Xu and Jan P. Loyen and Stefan M. Niehues and Moritz Augustin and Lennart Grosser and Marcus R. Makowski and Hugo JWL. Aerts and Alexander Löser},
journal= {arXiv preprint arXiv:2303.08179},
year = {2024}
}
Comments
Keno K. Bressem and Jens-Michalis Papaioannou and Paul Grundmann contributed equally