How to manually compute F-score, Precision and Recall

Hi there,

When a test is run on Rasa, a Confusion Matrix is generated. The f-scores, precision and recall are also automatically calculated.

Is it possible to manually calculate the above scores? If so how?

Hi Gabriel! Are you asking how they’re calculated? If so the Wikipedia article on them is actually really good](Precision and recall - Wikipedia) and that’s where I generally go when I forget the formulas ;). Or are you asking how to get the raw prediction data? We don’t currently have a batch prediction API but you can use the rasa nlu command line tool to get predictions item by item.

Hi Rachael,

Thanks for the wiki link. However, it’ll be the raw prediction I’m after.

Currently, when I test a test set against the training data, a confusion matrix is printed onto the command line. How can I be shown the data need to calculate the Precession, Recall and F-score?

Gabriel