MediumNLPnlp-178

Error Bucket Analysis

Problem

Bucket errors by their gold label: count the misclassified examples per gold label.

Return only labels that have at least one error, and an empty dictionary when there are none.

Hint

Key the buckets by the gold label, not the prediction.

case 1

in: [["a","b","c","a"],["a","c","c","b"]]

expected: {"b":1,"a":1}

case 2

in: [["a","b"],["a","b"]]

expected: {}

case 3

in: [[],[]]

expected: {}

case 4

in: [["x","y"],["y","x"]]

expected: {"x":1,"y":1}

Starter codePython

def error_bucket_analysis(gold_labels, predicted_labels):
    # Your code here
    pass

4 test cases · run your code against them in the editor