Weakly supervised contrastive representation learning to encode narrative viewpoint of COVID-19 tweets
Fig 2
UMAP latent space visualization of two baseline embedding methods (a and b) and a concatenated embedding combining DistilRoBERTa with node2vec embeddings (c) for a selected set of topics with manually annotated binary viewpoints.
(a) DistilRoBERTa. (b) CT-BERT. (c) DistilRoBERTa + n2v.