Span Corruption Ratio

Problem

Compute the fraction of tokens hidden by span corruption: the total length of the masked spans divided by the total number of tokens. Spans are given as [start, end) pairs.

The signature is span_corruption_ratio(masked_spans, total_tokens).

Hint

T5-style span corruption hides contiguous runs rather than individual tokens.

Test cases

case 1

in: [[[0,2],[5,7]],10]

expected: 0.4

case 2

in: [[[0,3]],3]

expected: 1

case 3

in: [[],5]

expected: 0

case 4

in: [[[1,2],[3,4],[5,6]],12]

expected: 0.25

Starter codePython

def span_corruption_ratio(masked_spans, total_tokens):
    # Your code here
    pass

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