Sistema baseado em software livre para reconhecimento de fala em nuvem em português brasileiro com alta disponibilidade

Abstract

This work aims to propose a solution containing an automatic speech recognition system in cloud. Thus, there is no need for a recognizer running on client machine itself since the same will be available via the Internet. In addition to cloud-based automatic speech recognition another aspect of this work is high availability. The importance of this topic is because the server environment where the recognition cloud will run must be available to the user. The various aspects that require robustness, such as Internet connection itself, the scope of this work was defined as free softwares that enable companies can use to increase the availability of their services. Among the results achieved under the simulated conditions, it was shown that the speech recognizer cloud developed by the group achieved a performance close to Google’s.