There's study here: http://cs230.stanford.edu/projects_spring_2021/reports/57.pdf http://cs230.stanford.edu/projects_spring_2021/reports/57.pd... Basically, yo

There's study here: http://cs230.stanford.edu/projects_spring_2021/reports/57.pdf http://cs230.stanford.edu/projects_spring_2021/reports/57.pd... Basically, you can collect text from different authors, then you can use authors names as labels to train a similarity learning with it. My suggestion would be finetune a Transformer model with a specific head and an ArcFace loss.