Enter a sentence in either English, German, or French, and the project will guess which language it is in. It is almost always accurate. This uses the PWSC algorithm and runs OFFLINE (no Google Translate). It is surprisingly accurate and almost a little scarily efficient for what it does. All of the three datasets come from the work of volunteers at the Tatoeba project. I sampled 1,000 sentences randomly in English, French, and German. I cleaned the English dataset already, but if there are any lines that should be removed in the French and German sets, please tell me.