Language and speech recognition method dynamically matching vocabulary to contents to be recognized, such as Internet sides

摘要

The data banks, made available by a server to a user, are examined by an algorithm for code interpretation for signs with respect to their link to other data banks. The obtained signs, i.e. words in a wider sense, serve then as an algorithm for speaker independent language recognition. The active vocabulary contains only a limited number of words to be recognized. If the algorithm cannot determine any connection between the input and the selected vocabulary, the user is instructed acoustically and/or visually to repeat his input by spelling, repeated speech, or click of keyboard or mouse.