HardNLPnlp-242

Chunk Overlap Coverage

Problem

Compute overlap statistics for sliding windows. Windows start at 0, stride, 2*stride, and so on while the start index is below n_tokens, and each window covers chunk_size tokens capped at n_tokens.

Return [n_chunks, covered] where covered is the number of distinct token positions covered. Return [0, 0] when any size argument is not positive.

Hint

The last window starts at ((n_tokens - 1) // stride) * stride and extends chunk_size positions.

Test cases

case 1

in: [10,4,2]

expected: [5,10]

case 2

in: [10,3,4]

expected: [3,10]

case 3

in: [5,2,3]

expected: [2,5]

case 4

in: [0,4,2]

expected: [0,0]

case 5

in: [4,6,2]

expected: [2,4]

Starter codePython

def chunk_overlap_coverage(n_tokens, chunk_size, stride):
    # Your code here
    pass

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