Run_evaluation: Entity Errors in NLU

Just like ‘run_evaluation’ saves intent errors in errors.json, is there way to get Entity Errors when model evaluation is done?

While ‘run_evaluation’ give precision, recall & F1 Score for entities extracted but it seems there is no straightforward way to figure out what caused bad precision, recall & F1 Score.

Currently, no. If you’d be willing to implement a feature like this, we’d be happy to accept a PR

Sorry, I won’t be able to. Though i have added it in issue just in case someone likes to pick it up.

1 Like