utilities to fix LaTeX issues

copienator/commands/post_correction.py

@@ -48,6 +48,15 @@ def escape_latex_underscores(text: str) -> str:
     return "".join(parts)
 
 
+def escape_latex_percents(text: str) -> str:
+    r"""Escape percent signs before annotation text is passed to LaTeX.
+
+    Unlike underscores, percent signs are comments both inside and outside
+    math mode.  Collapsing an existing escape run keeps cleanup idempotent.
+    """
+    return re.sub(r"\\*%", lambda _match: r"\%", text)
+
+
 def normalize_overescaped_latex_commands(text: str) -> str:
     r"""Collapse doubled command escapes inside LaTeX math environments.
 
@@ -118,7 +127,7 @@ def clean_string(text: str, lookup: dict[str, str]) -> str:
         text = fast_fix(text, lookup).replace("\x00", "")
     text = repair_json_escape_corruption(text)
     text = normalize_overescaped_latex_commands(text)
-    return escape_latex_underscores(text)
+    return escape_latex_percents(escape_latex_underscores(text))
 
 
 def clean_obj(value: Any, lookup: dict[str, str]) -> Any: