It’s important to understand these definitions as they form a key part of explaining other fundamental Machine Learning concepts like precision and recall.

 

The definitions below are outlined in the context of their application within the Re:infer platform.

 

To start with:

  • A ‘positive’ prediction is one where the model thinks that a label applies to a verbatim
  • A ‘negative’ prediction is one where the model thinks that a label does not apply to a verbatim

 

True positives

 

A true positive result is one where the model correctly predicts that a label applies to a verbatim.

 

True negatives

 

A true negative result is one where the model correctly predicts that a label does not apply to a verbatim.

 

False positives

 

A false positive result is one where the model incorrectly predicts that a label applies to a verbatim, when in fact it does not apply.

 

False negatives

 

A false negative result is one where the model incorrectly predicts that a label does not apply to a verbatim, when in fact it does apply.



To understand each of these concepts in more detail, please see here.



Previous: Validation     |     Next: Recall