Studying patterns of interest representation in politics is a central concern of scholars working on interest groups and lobbying. However, systematic empirical analysis of interest group representation entails a large amount of coding and is potentially prone to error. This letter addresses the potential of two computational methods in enabling large-scale analyses of interest group representation. We discuss the trade-offs associated with each method and empirically compare a manual, a query-based, and an off-the-shelf supervised machine learning approach to identify interest groups in a sample of 3000 news stories. Our results demonstrate the potential of automated methods, especially when used in combination.