Twitter has become an essential tool for candidates seeking election. It provides candidates with a direct line of communication to outside audiences and allows them to build a public profile by posting content that emphasises their electability and policy skills as politicians. Taking the 2020 Irish General Election as our case, this study seeks to examine how Twitter is used by candidates to signal campaign effort and policy priorities. In a series of experiments, we first demonstrate that a supervised machine-learning approach that incorporates word embeddings using transfer learning and employs data augmentations to extend our training set can successfully capture differences in how candidates present themselves online. We then explore how factors including electoral competition, incumbency, and gender drive these differences. Our results provide new insights into how multi-modal online communication strategies are used by candidates to build a public persona during electoral campaigns.