16 duplicated lines in:
  - mdr/qa/qa_dataset.py (17:35, 4%)
  - mdr/retrieval/data/data_utils.py (11:29, 9%)

11 duplicated lines in:
  - mdr/qa/hotpot_evaluate_v1.py (13:28, 10%)
  - mdr/retrieval/data/data_utils.py (31:46, 6%)