Linked Data provide many benefits to data consumers, but many publicly available datasets are still released in the Comma Separated Values (CSV) format, a ubiquitous common denominator. We introduce a methodology to transform such datasets into Linked Data. Our design is based on requirements identified while surveying existing governmental datasets released by data.gov. We present an implementation-independent RDF vocabulary to describe how a CSV dataset should be promoted into Linked Data, and use a Java-based converter to produce 5.3 billion RDF triples from 312 data.gov datasets.