​
Input

  • output (str): The generated text (set of items)
  • expectedOutput (str): The reference text (set of items)

​
Output

  • Result (float): A score between 0 and 1.

​
Interpretation

  • Higher scores (closer to 1): A larger fraction of actual positives are recovered (fewer false negatives)
  • Lower scores (closer to 0): Many actual positives were missed (more false negatives)

​
Formula

Recall=TPTP+FN\mathrm{Recall} = \frac{\mathrm{TP}}{\mathrm{TP} + \mathrm{FN}}
This is a Similarity Metric

​
Use Cases

  • Medical diagnosis and anomaly detection
  • Fraud detection
  • Any task where missing positives is costly