fix(question_gen): check_verbatim covers question_text + add missing blacklist patterns

- check_verbatim now computes n-gram overlap for BOTH question_text and
  correct_option vs source texts, returning max(question_ratio, option_ratio).
  Extracted _ngram_overlap_ratio helper for reuse.

- Added 4 missing blacklist patterns: 'this segment', 'this frame',
  'the current frame', 'frame summary'.

- Added 5 new test cases covering the above changes.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

app/question_gen/postprocess.py

@@ -52,6 +52,10 @@ _BLACKLIST_PATTERNS: list[tuple[re.Pattern[str], str]] = [
     (re.compile(r"the\s+clip", re.IGNORECASE), "the clip"),
     (re.compile(r"this\s+footage", re.IGNORECASE), "this footage"),
     (re.compile(r"the\s+footage", re.IGNORECASE), "the footage"),
+    (re.compile(r"this\s+segment", re.IGNORECASE), "this segment"),
+    (re.compile(r"this\s+frame", re.IGNORECASE), "this frame"),
+    (re.compile(r"the\s+current\s+frame", re.IGNORECASE), "the current frame"),
+    (re.compile(r"frame\s+summary", re.IGNORECASE), "frame summary"),
     (re.compile(r"上面的片段", re.IGNORECASE), "上面的片段"),
     (re.compile(r"这段视频", re.IGNORECASE), "这段视频"),
     (re.compile(r"该视频", re.IGNORECASE), "该视频"),
@@ -179,47 +183,56 @@ def check_referent_blacklist(question_text: str) -> list[str]:
     return violations
 
 
+def _ngram_overlap_ratio(
+    text_words: list[str],
+    source_ngrams: set[tuple[str, ...]],
+    window: int,
+) -> float:
+    """计算单段文本与来源 n-gram 集合的重叠率。
+
+    参数:
+        text_words: 待比较文本的分词列表（已小写化）。
+        source_ngrams: 来源素材的 n-gram 集合。
+        window: n-gram 窗口大小。
+
+    返回:
+        重叠率 [0.0, 1.0]，词数不足窗口大小时返回 0.0。
+    """
+    if len(text_words) < window:
+        return 0.0
+
+    text_ngrams: set[tuple[str, ...]] = {
+        tuple(text_words[i : i + window]) for i in range(len(text_words) - window + 1)
+    }
+
+    if not text_ngrams:
+        return 0.0
+
+    overlap = text_ngrams & source_ngrams
+    return len(overlap) / len(text_ngrams)
+
+
 def check_verbatim(
     question_text: str,
     correct_option: str,
     source_texts: list[str],
     window: int = 6,
 ) -> float:
-    """计算正确选项与来源素材的逐字重复率。
+    """计算题目文本和正确选项与来源素材的最大逐字重复率。
 
-    使用滑动窗口 n-gram 集合交集方法：从选项文本提取所有 n-gram，
-    与来源文本的 n-gram 集合求交集，计算重叠比例。
+    使用滑动窗口 n-gram 集合交集方法：分别从题目文本和选项文本提取 n-gram，
+    各自与来源文本的 n-gram 集合求交集，返回两者中较大的重叠比例。
 
     参数:
-        question_text: 题目文本（当前未使用，预留接口）。
+        question_text: 题目文本。
         correct_option: 正确选项文本（含 "X. " 前缀）。
         source_texts: 来源素材文本列表。
         window: n-gram 窗口大小。
 
     返回:
-        重复率 [0.0, 1.0]。0.0 表示无重叠，1.0 表示完全复制。
+        最大重复率 [0.0, 1.0]。0.0 表示无重叠，1.0 表示完全复制。
     """
-    # Phase 1: 提取选项纯文本（去掉可能的 "X. " 前缀）
-    option_text = correct_option
-    if len(option_text) >= 3 and option_text[1] == "." and option_text[2] == " ":
-        option_text = option_text[3:]
-
-    # Phase 2: 分词（简单空格分词，转小写）
-    option_words = option_text.lower().split()
-
-    # Phase 3: 选项词数不足窗口大小则无法构成 n-gram
-    if len(option_words) < window:
-        return 0.0
-
-    # Phase 4: 构造选项的 n-gram 集合
-    option_ngrams: set[tuple[str, ...]] = set()
-    for i in range(len(option_words) - window + 1):
-        option_ngrams.add(tuple(option_words[i : i + window]))
-
-    if not option_ngrams:
-        return 0.0
-
-    # Phase 5: 构造来源文本的 n-gram 集合
+    # Phase 1: 构造来源文本的 n-gram 集合
     source_ngrams: set[tuple[str, ...]] = set()
     for source in source_texts:
         words = source.lower().split()
@@ -229,9 +242,20 @@ def check_verbatim(
     if not source_ngrams:
         return 0.0
 
-    # Phase 6: 计算交集比例
-    overlap = option_ngrams & source_ngrams
-    return len(overlap) / len(option_ngrams)
+    # Phase 2: 计算题目文本的重叠率
+    question_words = question_text.lower().split()
+    question_ratio = _ngram_overlap_ratio(question_words, source_ngrams, window)
+
+    # Phase 3: 提取选项纯文本（去掉可能的 "X. " 前缀）
+    option_text = correct_option
+    if len(option_text) >= 3 and option_text[1] == "." and option_text[2] == " ":
+        option_text = option_text[3:]
+
+    option_words = option_text.lower().split()
+    option_ratio = _ngram_overlap_ratio(option_words, source_ngrams, window)
+
+    # Phase 4: 返回两者中较大的重叠率
+    return max(question_ratio, option_ratio)
 
 
 def has_time_anchor(question_text: str) -> bool:

tests/unit/test_postprocess.py

@@ -122,6 +122,31 @@ class TestReferentBlacklist:
         violations = check_referent_blacklist(q)
         assert len(violations) >= 1
 
+    def test_this_segment_caught(self) -> None:
+        """'this segment' 被检测为违规。"""
+        q = "What action is performed in this segment?"
+        violations = check_referent_blacklist(q)
+        assert len(violations) >= 1
+        assert any("this segment" in v.lower() for v in violations)
+
+    def test_this_frame_caught(self) -> None:
+        """'this frame' 被检测为违规。"""
+        q = "What object is visible in this frame?"
+        violations = check_referent_blacklist(q)
+        assert len(violations) >= 1
+
+    def test_current_frame_caught(self) -> None:
+        """'the current frame' 被检测为违规。"""
+        q = "Based on the current frame, what is happening?"
+        violations = check_referent_blacklist(q)
+        assert len(violations) >= 1
+
+    def test_frame_summary_caught(self) -> None:
+        """'frame summary' 被检测为违规。"""
+        q = "According to the frame summary, what is the main event?"
+        violations = check_referent_blacklist(q)
+        assert len(violations) >= 1
+
 
 # ---------------------------------------------------------------------------
 # TestVerbatim
@@ -186,6 +211,21 @@ class TestVerbatim:
         )
         assert ratio == 0.0
 
+    def test_question_text_overlap_detected(self) -> None:
+        """题目文本从来源复制但选项原创时，仍检测到高重叠率。"""
+        source = "The man walked slowly through the crowded market buying fresh vegetables"
+        # 题目直接复制来源的大部分内容
+        question = "The man walked slowly through the crowded market buying what?"
+        # 选项完全原创
+        option = "A. Fresh organic produce from the local farm stand"
+        ratio = check_verbatim(
+            question_text=question,
+            correct_option=option,
+            source_texts=[source],
+            window=6,
+        )
+        assert ratio > 0.0, "题目文本的逐字重叠应被检测到"
+
 
 # ---------------------------------------------------------------------------
 # TestTimeAnchor