The code of this project is archived to reflect the state of the PhD submission. Further development and refactoring can be found here.
- Execute the data/download_resources.sh script to download the models
- Depending on whether you want to use our trained Wikipedia word2vec models, you also need to download word embeddings via this script:
- If you want to use character embeddings from fasttext, you can download the specific model here: http://lager.cs.uni-duesseldorf.de/NLP/fasttext/german/wikipedia/de-wiki_20170501/
With the following script, you can download chracter (3, 3)-grams that were trained with 5 and with 50 iterations: