In this paper, we apply two methodologies of data enrichment to predict the case systems of languages from a diverse and complex data set. The methodologies are based on those of Bender et al. (2013), but we extend them to work with a new data format and apply them to a new dataset. In doing so, we explore the effects of noise and inconsistency on the proposed algorithms. Our analysis reveals assumptions in the previous work that do not hold up in less controlled data sets.
展开▼
