Ferramentas e recursos disponíveis para reconhecimento de fala em Português Brasileiro

RESUMO Speech recognition allows natural communication between the humans and machines. With Industry 4.0 there is a great demand for systems that perform this task, since human-machine integrations are increasingly attractive. Currently, there are several tools and resources that perform this activity, with some companies providing their audio recognition services through the Application Programming Interface, such as Microsoft, Google, IBM and Wit. On the other hand, there are offline libraries and open source that can also be explored like Vosk. Each company has its business rule and its specificity, in this sense it is difficult to know which is the most interesting for each situation. Thus, a comparison was made between speech recognition services in terms of usability, limitation and precision. In the comparison, speech recognition performance metrics were used in a set of audios, using the programming language Python.