This talk surveys and extends the dictionary model for identifying binding sites in non-coding regions of DNA. These sites control the transcription of genes into messenger RNA in preparation for translation into proteins. We summarize the underlying biology, review three different models for binding site identification, and present a unified model that borrows from the previous models and integrates their main features. We then describe maximum likelihood and maximum a posteriori algorithms for fitting the unified model to data. Time permitting, some specific examples will be discussed.