I should examine how different models for abstractive summarization generalize. The plan is to train them with different training data and observe how the summarizations produced are affected by this.
Unfortunately I found state of the art models hard to get working. Therefore I want to first test a simpler model which ideally well documented, easy to setup, (relatively) fast to train and can easily be trained with a different (sub)set of articles.
Is there such model you can recommend?
It does not matter if it is a few years old or produces bad summaries.