MediumNLPnlp-260

Grounded Answer Span

Problem

Locate answer_tokens inside doc_tokens as a contiguous subsequence. Return [start, end] inclusive indices of the first occurrence, or [-1, -1] when the answer is empty or absent.

Matching is exact and token-level.

Hint

Compare windows of the document with the full answer token list.

Test cases

case 1

in: [["the","cat","sat","down"],["cat","sat"]]

expected: [1,2]

case 2

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

expected: [-1,-1]

case 3

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

expected: [-1,-1]

case 4

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

expected: [-1,-1]

case 5

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

expected: [1,2]

Starter codePython

def grounded_answer_span(doc_tokens, answer_tokens):
    # Your code here
    pass

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