The 20 newsgroups dataset can be fetched through scikit-learn which fetches data from the 20 newsgroups website (http://qwone.com/~jason/20Newsgroups).
Embeddings are made with sentence BERT models where microsoft-mpnet-base (https://huggingface.co/microsoft/mpnet-base) is used as the base model.
For more details on models, see the model item: 10.11583/DTU.20708785
Funding
Danish Pioneer Centre for AI, DNRF grant number P1