Thanks! The post doesn’t cover fine tuning of the model which would be absolutely necessary (but out of scope for the post). Nils Reimers (the author of SBERT

Thanks! The post doesn’t cover fine tuning of the model which would be absolutely necessary (but out of scope for the post). Nils Reimers (the author of SBERT) has been on a speaking circuit covering Generative Pseudo Labelling to handle the vocabulary gap of new domains that a pretrained sbert model hasn’t seen yet. https://youtu.be/qzQPbIcQu9Q https://youtu.be/qzQPbIcQu9Q