Yes it is well known, but I wouldn't call it an issue per se, but rather a demonstration of the ambiguity of natural language. In any given language, many (most

Yes it is well known, but I wouldn't call it an issue per se, but rather a demonstration of the ambiguity of natural language. In any given language, many (most?) words have multiple meanings. Even if one of the underlying meanings of each word in your translation chain is "equivalent" (this could be a deep philosophical discussion of semantics on its own), each word-to-word correspondence for each language-pair has a different most-probable choice in the translation model (especially without context to help disambiguate).