Duolingo expects answers to be NFC-normalized strings (see the Unicode Normalization FAQ here: http://unicode.org/faq/normalization.html). But, depending on the input method, those values may contain decomposed characters, making them internally different from their normalized versions. For example, the Vietnamese word một (one) can be written in (at least) 3 different ways: một : m \u1ED9 t (NFC) một : m \u00F4 \u0323 t (VN keyboard) một : mo \u0323 \u0302 t (NFD) The extension aims to prevent false positives in typo detection by automatically normalizing relevant inputs.