I’m just back from INLG 2019 in Tokyo, where I was very happy to see an increased emphasis on evaluation (and other methodological issues), including several papers on improving human evaluations.
Advice on how to evaluate an NLG system by asking human subjects to rate the system, when they are using it in a real-world context.
A high-level discussion of the different ways of evaluating NLG systems.
Advice on how to evaluate an NLG system by asking human subjects to rate the system, in an artificial experiment context (ie, not real world usage). This is the most common type of NLG evaluation
