Key points are not available for this paper at this time.
This paper presents the machine learning architecture of the Snips Voice, a software solution to perform Spoken Language Understanding on typical of IoT devices. The embedded inference is fast and while enforcing privacy by design, as no personal user data is ever. Focusing on Automatic Speech Recognition and Natural Language, we detail our approach to training high-performance Machine models that are small enough to run in real-time on small devices. , we describe a data generation procedure that provides sufficient, -quality training data without compromising user privacy.
Coucke et al. (Fri,) studied this question.