Speech Recognition Swiss German

Bachelor Thesis Lukas Zimmermann 2016, see project page

Speech recognition with Swiss German is still a relatively new territory. Especially in the field of advisory settings, there is a lot to be done. This project should make it possible to identify various technical terms in such a conversation and to record them.

Using so-called Hidden Markov Models (HMM) allows to detect some predefined terms from the consulting environment. It is shown how much effort and how many data are required to create functioning speech recognition applications. In addition the project explains what frameworks are suitable for this problem. The resulting concept is a starting point for further research.