Problem

It may be undesirable for translators or MT/LLM engines to add punctuation marks to the translation that do not appear in the source, e.g., adding brackets or a colon.

The reverse may also be undesirable: punctuation marks used in the source, such as brackets or a colon, may be omitted from the translation.

LanguageTool can only detect punctuation errors within a single language, such as unpaired brackets. However, it cannot compare the punctuation used in the source and target texts.

Solution

Add a check for punctuation marks that have been added or omitted:

Source–target punctuation mismatch

• Punctuation mark added to the target
• Punctuation mark missing from the target

The check must identify all punctuation marks in the source and target texts, compare them, and report any discrepancies. For this, a definitive list of the punctuation marks covered by the check must be established as part of the implementation. This list should specify which characters and language-specific variants are recognized and compared if at all. The check can be language pair agnostic, except in the following cases (more to add):

Special languages French, Spanish:

French

For French, a non-breaking space is required before the following punctuation marks: colon :, semicolon ;, question mark ?, and exclamation mark !.

For example, if the source contains a colon, the expected sequence in the French target is a non-breaking space followed by a colon.

The check must therefore treat each non-breaking space and its associated punctuation mark as a single language-specific pair. This pairing must be considered when detecting and evaluating punctuation mismatches whenever French is the source or target language.

Spanish

Spanish uses an inverted question mark (¿) or exclamation mark (¡) at the beginning of an interrogative or exclamatory clause, in addition to the corresponding closing mark (? or !).

Example: Si vienes, ¿a qué hora llegarás?
(If you come, what time will you arrive?)

When Spanish is either the source or target language, the check must recognize ¿…? and ¡…! as language-specific punctuation pairs. The opening inverted mark must not be reported as an addition, and its absence from a Spanish text must be reported as a missing punctuation mark.