fix(regulatory): 修复换行产品名称提取不全

review_agent/regulatory_review/services/text_extract.py

@@ -21,6 +21,7 @@ class ExtractedText:
 
 
 SUPPORTED_EXTENSIONS = {".txt", ".md", ".pdf", ".docx", ".pptx", ".xlsx", ".doc"}
+FIELD_LABELS = ["产品名称", "型号规格", "预期用途", "管理类别", "分类编码", "注册类型", "临床评价路径"]
 
 
 def extract_text(path: str | Path) -> ExtractedText:
@@ -69,8 +70,32 @@ def _section_candidates(text: str) -> list[str]:
 
 def _field_candidates(text: str) -> dict[str, str]:
     fields = {}
-    for label in ["产品名称", "型号规格", "预期用途", "管理类别", "分类编码", "注册类型", "临床评价路径"]:
-        match = re.search(rf"{label}[:：]\s*([^\n\r]+)", text)
-        if match:
-            fields[label] = " ".join(match.group(1).strip().split())
+    lines = text.splitlines()
+    for index, line in enumerate(lines):
+        normalized = line.strip()
+        if not normalized:
+            continue
+        for label in FIELD_LABELS:
+            match = re.match(rf"^{re.escape(label)}[:：]\s*(.*)$", normalized)
+            if not match or label in fields:
+                continue
+            value_parts = [match.group(1).strip()]
+            for next_line in lines[index + 1 :]:
+                continuation = next_line.strip()
+                if not continuation or _starts_field_line(continuation) or _looks_like_section_heading(continuation):
+                    break
+                value_parts.append(continuation)
+            value = " ".join(part for part in value_parts if part)
+            if value:
+                fields[label] = " ".join(value.split())
     return fields
+
+
+def _starts_field_line(line: str) -> bool:
+    if any(re.match(rf"^{re.escape(label)}[:：]", line) for label in FIELD_LABELS):
+        return True
+    return bool(re.match(r"^[^\s:：]{2,24}[:：]", line))
+
+
+def _looks_like_section_heading(line: str) -> bool:
+    return bool(re.match(r"^([一二三四五六七八九十]+[、.．]|[0-9]+(\.[0-9]+)*[、.．\s])", line))