Labial reduplication in Hungarian is a semi-productive morphological template with a diminutive/ hypocoristic function, which consists in creating a copy of the base with a labial consonant in initial position; some examples are cica-mica ‘cat.DIM’, csiga-biga ‘snail.DIM’, Julcsi-bulcsi, hypocorism for Julia. While there are detailed accounts of similar phenomena in other languages, the Hungarian pattern has not been systematically described before. The lack of any useful descriptive material is partly due to the fact that labial reduplication exhibits a high degree of variability and is difficult to tease apart from a more general tendency to form rhyming compounds. In this talk, I analyse data collected from a 1.5 billion word corpus of Hungarian and attempt to isolate and describe this pattern with the help of machine learning algorithms and statistic methods. I also discuss some implications of the data for theories of morphology and phonology.