A recent trend in the evolution of the Web is the massive contribution of small chunks of content by regular users, typically in combination with mechanisms for fostering social interaction. Such is often referred to as microblogging, with Twitter, Identi.ca, and Google Buzz being the most widely known services. In this paper, we propose to use the underlying interaction pattern, and existing respective services, for the collaborative construction and maintenance of structured knowledge representations. We define (1) a syntax for embedding triple-like statements in Twitter messages, (2) develop a transformation into RDF, (3) suggest mechanisms for controlling the inclusion of such statements made by other users, (4) exploit the resulting graph for query expansion, and thereby provide a direct incentive for users to adopt our syntax; and (5) demonstrate the approach by means of a prototype implementation. The resulting RDF graphs can be combined easily with other Semantic Web data.