For the substantial analysis conducted in this book, parties’ policy positions at two stages of the representation process must be identified. This is achieved through an automated content analysis comprising two steps. This chapter describes the first step, in which the political texts of parties—manifestos and plenary speeches—are classified into policy domains. The policy domain classification is conducted using supervised learning methods. The units of analysis are manifesto sections and individual plenary speeches delivered by MPs and ministers. The policy domains for classification are defined by ministerial portfolios. This enables the simple creation of a vast set of training data for building the algorithm, namely through speeches by ministers in parliament. The trained algorithm assigns each text to one of up to 16 policy domains; the exact number and framing of these policy domains depend on the legislative term from which the text originates. The chapter concludes with a detailed evaluation of the accuracy and validity of the policy domain classification.