Update predictor.py

predictor.py

@@ -155,6 +155,46 @@ class SentenceExtractor:
         keyword_clean = keyword.translate(trans)
         return keyword_clean in sentence_clean
 
+    def _is_negated_positive(self, text: str, keyword: str) -> bool:
+        """检测积极关键词是否被否定词修饰，例如：
+        - 没有/无/不/非/未/并不/毫无 + 关键词
+        - 对以“有”开头的积极词（如“有创新性”），也匹配“没有/无/不/未/并不/毫无 + 去掉‘有’后的部分（如“创新性”）”
+        - 缺乏/不足/欠缺/缺少/不具备 + 关键词 或 关键词去“有”后的部分
+        """
+        if not keyword:
+            return False
+
+        sentence = text.strip()
+        neg_prefixes = [
+            "没有", "没", "无", "不", "非", "未", "并不", "并没有", "并无", "毫无"
+        ]
+        lack_prefixes = [
+            "缺乏", "不足", "欠缺", "缺少", "不具备", "不够"
+        ]
+
+        # 构建安全的正则片段
+        import re
+        def any_prefix(prefixes: List[str]) -> str:
+            return "(?:" + "|".join(re.escape(p) for p in prefixes) + ")"
+
+        patterns: List[str] = []
+        # 直接：否定前缀 + 关键词
+        patterns.append(rf"{any_prefix(neg_prefixes)}\s*{re.escape(keyword)}")
+
+        # 直接：缺乏类前缀 + 关键词
+        patterns.append(rf"{any_prefix(lack_prefixes)}\s*{re.escape(keyword)}")
+
+        # 若积极词以“有”开头，额外匹配去掉“有”的尾部（例如 ‘有创新性’ → ‘创新性’）
+        if keyword.startswith("有") and len(keyword) > 1:
+            tail = keyword[1:]
+            patterns.append(rf"{any_prefix(neg_prefixes)}\s*{re.escape(tail)}")
+            patterns.append(rf"{any_prefix(lack_prefixes)}\s*{re.escape(tail)}")
+
+        for pat in patterns:
+            if re.search(pat, sentence):
+                return True
+        return False
+
     def _extract_relevant_sentences(self, text: str) -> List[str]:
         sentences = self._split_into_sentences(text)
         relevant_sentences = []
@@ -188,8 +228,13 @@ class SentenceExtractor:
                 continue
             for keyword in self.eval_keywords[category].get("positive", []):
                 if self._fuzzy_match_keyword(text, keyword):
-                    positive_count += 1
-                    total_score += 1
+                    # 遇到被否定的积极词（如“没有创新性”含“有创新性”），按消极计分
+                    if self._is_negated_positive(text, keyword):
+                        negative_count += 1
+                        total_score -= 1
+                    else:
+                        positive_count += 1
+                        total_score += 1
             for keyword in self.eval_keywords[category].get("negative", []):
                 if self._fuzzy_match_keyword(text, keyword):
                     negative_count += 1
@@ -248,7 +293,3 @@ class SentenceExtractor:
         }
 
 
-
-
-
-