File size: 1,462 Bytes
f157ecc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 | """Minimal answer cleanup for Medico 2026 Task 1 submissions.
Full-test rescoring showed that semantic or qtype-specific normalization can
reduce the official text metrics. Keep this module intentionally conservative:
only remove obvious formatting noise and never rewrite the medical meaning.
"""
from __future__ import annotations
import re
_SPACE_RE = re.compile(r"\s+")
_LEADING_ANSWER_RE = re.compile(r"^\s*(?:answer\s*[:\-]\s*)+", re.IGNORECASE)
def compact_spaces(text: str) -> str:
return _SPACE_RE.sub(" ", str(text or "")).strip()
def normalize_prediction(prediction: str, question: str = "") -> str:
"""Return a minimally cleaned prediction without semantic rewrites."""
del question # The final Task 1 normalizer is intentionally question-agnostic.
text = compact_spaces(prediction)
text = _LEADING_ANSWER_RE.sub("", text)
text = re.sub(r"\s+([,.;:!?])", r"\1", text)
return compact_spaces(text).strip()
def infer_question_type_for_normalization(question: str) -> str:
"""Compatibility shim for older imports; not used for rewriting."""
q = str(question or "").lower()
if any(marker in q for marker in ["how many", "number of", "count"]):
return "numerical_count"
if q.startswith(("is ", "are ", "was ", "were ", "do ", "does ", "did ", "can ", "has ", "have ")):
return "yes_no"
return "other"
__all__ = ["normalize_prediction", "infer_question_type_for_normalization"]
|