Texts such as political speeches, social media posts, newspaper articles and open-ended survey responses convey a wide range of important information for social scientists. They are now available in digitised form on an unprecedented scale, allowing us to turn texts into data and study them systematically. This lecture will introduce the basic principles of computational text analysis, including its key advantages and disadvantages; the assumptions that lie behind it; the workflow of a typical text analysis project; the distinction between supervised and unsupervised approaches; a brief introduction to some core methods in the field including dictionary analysis, classification, topic modelling and textual scaling. Examples and applications will be drawn from the study of the politics of social policy, including the speaker’s own work on measuring discourse about welfare and its users from political speeches and newspaper articles. The talk will stress the opportunities that computational text analysis offers for scholars but also its difficulties and limitations, and the choice between qualitative and quantitative approaches.