"""Keyword coverage scorer for subjective answers.
中文：主观题评分规则层，基于关键词覆盖率计算得分与反馈（不含框架/路由逻辑）。
"""
from __future__ import annotations
from typing import Dict, List, Sequence
from ...pydantic_models import GradeAnswerRequest, GradeAnswerResponse
def _tokenize(text: str) -> List[str]:
return [
token
for token in text.lower().replace("，", " ").replace("。", " ").split()
if token
]
def _keyword_stats(std_tokens: Sequence[str], stu_tokens: Sequence[str]) -> Dict[str, List[str]]:
std_set = set(std_tokens)
stu_set = set(stu_tokens)
overlap = sorted(std_set.intersection(stu_set))
missing = sorted(std_set - set(overlap))
return {"matched": overlap, "missing": missing}
def score_answer(payload: GradeAnswerRequest) -> GradeAnswerResponse:
std_tokens = _tokenize(payload.correct_answer)
stu_tokens = _tokenize(payload.student_answer)
stats = _keyword_stats(std_tokens, stu_tokens)
coverage = len(stats["matched"]) / max(len(std_tokens), 1)
similarity = round(coverage, 3)
if not payload.student_answer.strip():
feedback = "系统未检测到作答，可提醒学生先尝试回答。"
elif similarity >= 0.9:
feedback = "作答内容非常完整，与标准答案高度一致。"
elif similarity >= 0.7:
feedback = "作答基本正确，可以补充缺失关键词以更全面。"
elif similarity >= 0.4:
feedback = "部分要点正确，建议强化缺失的概念。"
else:
feedback = "与标准答案差距较大，可引导学生重新审题并列出要点。"
return GradeAnswerResponse(
is_correct=similarity >= 0.7,
similarity=similarity,
matched_keywords=stats["matched"],
missing_keywords=stats["missing"],
feedback=feedback,
)
