This disclosure describes techniques to automatically detect a user’s proficiency in particular languages. Signals from user activity, such as search, browsing activity, location history, mobile usage, etc. are interpreted as indicators of the user’s likely language proficiency. Further, user preferences, such as language preferences, spell checking preferences, etc. are analyzed as indicators of the user’s language proficiency. Multiple signals are combined to determine languages in which the user is proficient. Detected proficiency is utilized to customize user experience, e.g., to automatically offer translation for content that is in a language in which the user is not proficient.
Blum, Rachel; Liutikas, Aurimas; Ainslie, Alex; and Simpson, Rachel, "Automatic Detection of User Language", Technical Disclosure Commons, (February 04, 2016)