Using pre-trained BERT embeddings for multi-class text classification

What would be the steps involved in doing such a thing? Basically I have data from a research,
the form of a dataframe including: participant ID, response ID, single word object, and response, which consists of a description of the function of the single word object. Some paricipants have had more than one response from the same object. Object is the same in the whole dataset, and category codings have been done by someone else. Basically responses need to be …

bert classification languagetechnology text text classification