I have a data set with a document-summary pair and I want to train it on something like XLNet to do abstractive summarization. I’m not completely versed in machine learning to develop it from scratch and was wondering how I would go about this? Would it be something similar to a translation task?