Automatic morphological processing can aid downstream natural language\nprocessing applications, especially for low-resource languages, and assist\nlanguage documentation efforts for endangered languages. Having long been\nmultilingual, the field of computational morphology is increasingly moving\ntowards approaches suitable for languages with minimal or no annotated\nresources. First, we survey recent developments in computational morphology\nwith a focus on low-resource languages. Second, we argue that the field is\nready to tackle the logical next challenge: understanding a language's\nmorphology from raw text alone. We perform an empirical study on a truly\nunsupervised version of the paradigm completion task and show that, while\nexisting state-of-the-art models bridged by two newly proposed models we devise\nperform reasonably, there is still much room for improvement. The stakes are\nhigh: solving this task will increase the language coverage of morphological\nresources by a number of magnitudes.\n