MTG
Automatic translation of sound descriptions
Sound descriptions in English sometimes become a problem for non-english speakers. A typical case is in educational contexts like in sound recording workshops at schools, but that is a rather general problem. The Freesound policy is that typically we'll accept descriptions in languages other than English as long as there are at least some tags (and ideally some shortened version of the descriptor) in English. This is to make sure sound is still findable and "understandable" for everyone.
A solution to this problem is to use automated translations from whatever language to English and show also the translated version of sound descriptions along with the original (properly marked as "automatic translation"). Also, at sound indexing time, a number of known keywords can be extracted from the translation and used for indexing (or the full translated text).
To do that we'd probably need to use some external API that detects original language and provides a translation. Then we'd need to have a translated_description field in the Sound model to store the translated version. translated_description should be re-computed every time sound description is updated (which almost never happens). We should also think what to do with tags and filenames. We probably don't want to translate filename but want to show translated tags.
This is a very broad approximation to how something like this could word, but can serve as starting point for a discussion.
