Q1. Thinking about a confidence score (that tries to predict the accuracy of a prediction) in OpenNMT-py, what do you all use?
Q = PRED SCORE / len(target sentence)
Or what?
Q2. Has anyone plotted this or any other Q vs the ultimate BLEU?
Q1. Thinking about a confidence score (that tries to predict the accuracy of a prediction) in OpenNMT-py, what do you all use?
Q = PRED SCORE / len(target sentence)
Or what?
Q2. Has anyone plotted this or any other Q vs the ultimate BLEU?