With proliferation of the Internet of Things, annotation and generation of metadata describing data streams produced by sensors becomes even more urgent and important. This article proposes a method of annotating data streams with voice and extracting semantics from data. The strengths and weaknesses of existing voice recognition systems are discussed and it is argued that ontologies should play important role in making annotations meaningful and useful for various services and applications, including annotating road conditions and traffic situations. The architecture and implementation of the proposed system is discussed and demonstrated.