Difficulties for Machine Translation Systems

2.4.1 Ambiguous Words and Structures
Most translation systems with their piece of artificial intelligence are able to find out the subject and even a rough idea of the text. Often they can associate the text with a vocabulary of a special topic. This helps deciding between different meanings of words with the same spelling. For example "bank" is a homonym because it has two unrelated meanings. Difficulties also occur because of "multiple meanings" polysemes have.  E.g. "line" can stand for a limit, a row or a string, meanings from the same etymology. Both homonym and polyseme translations require comprehension because the meanings depend on the context. The subject detection seems to be a useful solution although in a few cases bugs will remain if the topics are not clearly distinguishable. The following German example demonstrates how a sentence can lead to a wrong deduction: "
Sie spielt 'Amsel, Drossel, Fink und Star' auf dem Flügel." The system may find out that the sentence is in relation with birds. So there is no reason for using artistic vocabulary instead of natural (animal) vocabulary. That's why the translation for "Flügel" wrongly outputs "wing" in this case. Using webtranslate you get "She plays 'blackbird, thrush, finch and starling' on the wing.". If one tries "Sie spielt ein Lied auf dem Flügel", the translation "She plays a song on the grand piano" is fine because “Lied” is a clear hint for artistic vocabulary. The rare Example-Based Translation Systems often benefit because they have, for example, a dictionary entry for "auf dem Flügel spielen" which tells them to tend to translate this combination of words with "play the grand piano". That’s why they are not that often under a delusion.