The analysis of your model performance can be done by clicking on the « Result » button in the « Model » section. The results are only available if at least one test set has been added when creating the model.
Choose your test set
If you have added more than one test set to your model, you can analyze the results of your model against these different test sets by clicking on the top left dropdown button.
Choose your NER
If you have added more than one NER component to your training pipeline, you can analyze the results of your model for each NER by clicking on the extractor dropdown menu.
The analytics section gathers in one single place the different performance indicators of your model : recall, precision, F1-score.
These indicators are available globally and intent by intent.
The confusion matrix enables you to visualize quickly the behavior of your model intent by intent.
The top error widget enables you to have a detailed understanding of your model’s NER main errors:
- Click on « Added values » to see all the values added by your model
- Click on « Missed values » to see all the values missed by your model