Journal Club: Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

To measure the semantic similarity of text strings and address the computational efficiency of using BERT, Reimers and Gurevych developed a model that uses pretrained BERT models in a siamese network structure to create sentence embeddings. They evaluated SBERT on common semantic textual similarity tasks and assessed its computational efficiency. Here are Google Slides with a journal club presentation: Journal Club: Sentence-BERT

Reimers and Gurevych 2019: https://arxiv.org/abs/1908.10084